History Home | Book Catalog | International Catalog of Sources | Visual Archives | Contact Us

Oral History Transcript — Dr. Syukuro Manabe

This transcript may not be quoted, reproduced or redistributed in whole or in part by any means except with the written permission of the American Institute of Physics.

This transcript is based on a tape-recorded interview deposited at the Center for History of Physics of the American Institute of Physics. The AIP's interviews have generally been transcribed from tape, edited by the interviewer for clarity, and then further edited by the interviewee. If this interview is important to you, you should consult earlier versions of the transcript or listen to the original tape. For many interviews, the AIP retains substantial files with further information about the interviewee and the interview itself. Please contact us for information about accessing these materials.

Please bear in mind that: 1) This material is a transcript of the spoken word rather than a literary product; 2) An interview must be read with the awareness that different people's memories about an event will often differ, and that memories can change with time for many reasons including subsequent experiences, interactions with others, and one's feelings about an event. Disclaimer: This transcript was scanned from a typescript, introducing occasional spelling errors. The original typescript is available.

Access form   |   Project support   |   How to cite   |   Print this page


See the catalog record for this interview and search for other interviews in our collection


Interview with Dr. Syukuro Manabe
By Paul Edwards
At Princeton University
March 15, 1998

open tab View abstract

Syukuro Manabe; March 15, 1998

ABSTRACT: In this interview, Syukuro Manabe discusses: University of Tokyo; computer modeling; National Oceanic and Atmospheric Administration (NOAA); greenhouse effect; Geophysical Fluid Dynamics Laboratory (GFDL); Intergovernmental Panel on Climate Change (IPCC). Individuals discussed include: Joe Smagorinsky; Kikuro Miyakoda; Akio Arakawa; Jule Charney; John Von Neumann; L. F. Richardson; Douglas Lilly; Bob Strickler; Leith Holloway; Giichi Yamamoto; Sigmund Fritz; Wally Broecker; Sydney Levitus.

Transcript

Session I | Session II

Edwards:

Great. Well, here we are again for Day 2 of your oral history.

Manabe:

Yeah. Now the one thing I thought I would make clear from yesterday's thing is that, so what happened is sort of by the early part of 1970 we had our, constructed our models and found out that the model is working very well beyond the expectation in simulating hydrologic cycle, [???] general circulations, stratosphere, which is not as good but still it's pretty decent stratospheric circulations. And so that in a sense that what Joe dreamed of when he started this, our system science model there in 1950, gradually sort of come into place. And then we started putting in additions, not only putting radiative transfer in the hydrologic cycle to this original GCM, but also begin to put the effect of ocean into the thing, so that that is sort of around early 1970 which we felt —

Edwards:

This is your work with Kirk Bryan [?].

Manabe:

That [???] ocean one was with Kirk Bryan. We seemed to have achieved these things. And basic principle of constructing these models is that if we know the process of basic fundamental law of physics, then we will put it explicitly in, including hydrodynamics, which I don't claim that much myself for contributing that much. Other people did, and I used it I think. And then other things —

Edwards:

Who were the most important people in contributing to the hydrodynamic side?

Manabe:

Hydrodynamics, I think the people who did, the pioneers of numerical weather prediction. So Jules Charney, Norman Phillips, Art Arieson [?], and these are the pioneers. And in terms of numerics of course Arakawa did excellent work on it. And so these are the people, the pioneers who did the dynamics. And also though there is other physical process, which we know very well by then. That is a radiative transfer. And so I decided to put that explicitly and a straightforward way rather than using sort of highly parameterized form, which everybody else was doing. And then we also have, the most important thing in the hydrologic cycle is Clausis Clapayron Equation, what is the saturation vapor pressure depends upon temperature.

Edwards:

Okay. What did you call it?

Manabe:

Clausis, C-l-a-u-s-i-s. Clapayron, C-l-a-p-a-y-r-o-n. And this makes some dynamical role which decides when air gets saturated how much moisture you have when the air gets circulated. And that is a most important role is modeling hydrologic cycle. And so then we know dynamics, radiative transfer, and the law of thermodynamics on circulation vapor pressure. And these are the things, and probably simple heat and water balance on the land surface, that first dynamics, first law of thermodynamics. We put these things explicitly in a rather straightforward way, but we tried to do some of the other things as simple as possible, like cumulus convection. I did it with simple convective adjustment, and the land surface process which is also very difficult if you try to incorporate heterogeneity of land surface and how to aggregate this degreed scale. It's very complicated. We put a very simple bucket model. So the basic thing is if you know the laws of physics you put explicitly, as faithfully as to the law of physics. However, if you don't know enough you will make as simple as possible. And so that is more or less my basic principles, and don't collect [?] a model around with the parameterization of physics which we don't understand satisfactorily.

Edwards:

Now how would you say about Arakawa's work on cumulus parameterization in respect to what you just said? Because your method for handling that was a simple convective process.

Manabe:

Very simple. Yeah.

Edwards:

His became quite complex as time went on.

Manabe:

Yeah. That's right. And so I will get into that. And so what happened is then what I found by the early 1970, despite all these simplifications, I get beautiful hydrologic cycle, distribution of global rainforests and so that what I felt was that this is good enough to do the areas, application of these models for various things. That is early 1970. So I started putting this application of atmospheric GCM and a coupled GCM, starting from middle 1970.

Edwards:

Right.

Manabe:

And that's the time when all the other people started pushing more into improving parameterizations. And this explains why I didn't try to incorporate. Because when you want to use it, worst thing you can do is to keep changing the model. And so during, I told you, the second half of my career was application of these models for various problems — anthropogenic change, natural variability climate change, apply these models to these things. And so that was what I was doing. And in many of the things that are like cumulus parameterization, which Arakawa initiated and then many people have been improving on it, is therefore major activity occurred after I made up my mind to —

Edwards:

I see.

Manabe:

So that is the historical sort of account of what happened what. And my reluctance to put many of the complexities in is coming from partly attributable to that, my basic — And but another thing however is that the cumulus parameterization, there is two parts. One is the dynamical closure [?], what he called mass flux and how air rises in the condensation [???], how it sinks and what the net effect on the large scale temperature and humidity field. If you have this kind of, many, many of these things. And but then really the sort of — another part which played a very important role in the cumulus parameterization. That is cloud microphysics. So [???] air may not go up and down like this, but right next to up draft there may be down draft. Very intense downward motion and droplets evaporating as it comes down. Or it may not go this way, it may go this way, but above all so that the cloud microphysics is much less known, so that even though you make a very sophisticated program of dynamical part, microphysical part we know very little. And —

Edwards:

This has been Richard Lindzen's criticism of climate models for many years.

Manabe:

[???]? Yeah. So that is one problem. And then there is another problem. That is in the statistical mechanics or something in the law of physics. You try to explain law of thermal dynamics by statistical mechanics. So you have idealized model of molecules, and then when you put many of them, what did microscopic behavior matters. Now, and then try to explain thermodynamical law. Now, in this case there is a difference in scale between molecules and microscopic thermodynamics. It's a huge difference. However in the parameterization of cumulus convection, there is a small cumulus which occurs something like Eastern Pacific or somewhere, but there is a much bigger cumulonimbus all the way to mesoscale convective cloud or a large scale convective cloud which is the entire largest scale cyclone, part of cyclone, is convecting up and down all of these complicated things. It's shrouded with cloud. So what is difficult about it is there is no spectral gap between large scale and small scale [???].

Edwards:

Spectral gap?

Manabe:

Yeah. That is, if you can say this is the small scale, you are parameterizing to describe the effect of small scale on large scale. Okay. If you can do that, it would be very nice. The trouble is, you don't know where to cut and say this is a small scale; this is a large scale because there is a continuous spectrum of this scale all the way from large scale to small scale. So you can't arbitrarily cut this thing and then say this large scale, this is small scale. And, you know, [???] as agreed scale becomes smaller and smaller it's getting even more difficult, because convective scales are splitting artificially by grid scale. And so these are some of the, so that what I would say is that it's easy to say I can parameterize this, I am going to parameterize that. But in practice it is a very, very difficult problem. The tolerance, theory of tolerance, all they have sold [?] until now is idealize the situation or homogeneous isotropic turbulence theories available.

Edwards:

Isotropic what?

Manabe:

Turbulence. T-u-r-b-u-l-e-n-c-e.

Edwards:

Okay.

Manabe:

Yeah. So, that is always so. But here we are talking about the, well the eddies which had a continuous spectrum, and then on top of it it's convecting. Air is going up and down, it's not homogeneous isotropic. Okay. And then furthermore there is a cloud microphysics put on top of it, and then this water droplet rising and falling and interacting with turbulence. Okay. So you can see the degree of this challenge. So I was standing in his [???] of Arakawa's. You know, he picked up most difficult problem in atmospheric science — which is literally true.

Edwards:

Yes.

Manabe:

Right? What I did was I pumped it [punted?].

Edwards:

[laughs]

Manabe:

Okay? And then put a convective adjustment in and then I said all [???] scale, all it does is eliminate static instability, so that I don't have to deal with this having to resolve this sub grid scale, most unstable convection by large scale, which is impossible. So I just eliminated unstable stratifications and solved the problem at that time. Now if I try to solve that in early 1960, I would be still doing it as Arakawa has been doing. And you can see how difficult this is. And so that the same thing can be said about land surface process. Now you look at, I put this simple bucket so if rain falls you put the water in, if it evaporates you take water out, and if it rains more than evaporation you fill the box, it overflows the box. And that's become river runoff, and flow into ocean and dilutes the salinity of the ocean water. So I had that very simple system then. And that's, but if you try to make it complicated, then you have one tree and leaves, and this leaves, some of the planetary physiologists' parameterization, many of them are like that. Leaves are as big as grid size, so if it's a 200 x 200 kilometer, 200 kilometer leaves. And then there is [???] resistance, the sun shines about there, and water evaporates from the leaves as well as soil surface. Okay? And then you begin to see how the water is sucked out of the root system. So it is easy to put many, many difference layers into soil, but how to parameterize the sucking of water through root system. Okay. But then that is planet [plant?] physiologist's view of land surface process. But then there is another one which is hydrologist's view. Hydrologists realize inside one grid box there is many catchment [?] in the river. They have the sloping surface and water flows along, and then go to each catchment, and then eventually go into river. So hydrologists are much more interested in how to aggregate individual trees and pastures and sandy soil region and which exist inside each grid box, how to aggregate them to create the large scale behavior of water and heat balance. And either one of these two groups are capable of making the model extremely complicated by their own way of either through aggregation or either through plant physiology.

Edwards:

Right.

Manabe:

Okay. And so that, and if you put these two things together you got a complex square. And furthermore it is not clear if you have some, you know you see some burned soil [?] in the sky, and they are looking at the thermos [?] which air is rising. So they found there the [???]. So inside the grid box there is many thermals there. So it's not maybe linear averaging of each heterogeneous thing inside, but they may be nonlinearly interacting inside. Different part of the sub grid scale area may be interacting with each other through atmospheric circulations. Okay. So that — which cannot be treated by this cross-grid models of course. But if you want to make a grid fine, how fine you have to. And again, problem exists that no spectral gaps between. Everything small scale to large scale things continuously evolve, interact at different scale. And so again I pump it. And then put this simple box in. And what I found in early, no, late 1960, now that I have a global model and put these things in, were late 1960 and to early '70 is the simulations — simulated distribution of rain, circulation which unfold in front of my eyes is unbelievably good. At least not in a regional scale, but in a large global sense, so that wherever there is a desert hardly any rain is falling. Wherever there is a tropical rain belt, model has simulated beautifully the tropical rain belt. Almost all desert is simulated—Sahara Desert, Southwestern Desert, Gobi Desert, which is not quite perfect but it's some simulations, and the Patagonian Desert, Kalahari Desert, you name it, every one of these arid regions are reproduced in the model. And so that I was very happy that despite all these simplifications. But then most of people said, "Oh, got it for wrong reason." I thought that these things are, rainfall is controlled by large scale dynamics, how much air converted in the cyclone waves. And so if we get reasonable simulation of cyclone waves. Or in the tropics it's controlled by sea surface temperature, and so that if we put right sea surface temperatures [???] recognition [?] we could get right rainfall, induce right convergence and so forth. So that essentially it's a large scale process. Doesn't depend upon glorious [?] gaudy [?] details of parameterizations. And I felt I am sort of vindicated by my basic approach to make things which we don't know enough, to make it as simple as possible, and then what we know enough, we treat them explicitly. And so that was the basic philosophies, and because its simulation was not perfect on the other hand, but I thought this is good enough to do, study wide variety of problems. And so I decided to. And meanwhile other people probably saw how good my simulation was. Although they don't cite it much more, because I published too early. Okay. And but then they said okay, this could be improved even further. So they started putting the areas of improved parameterization, and many people in a different discipline, the people in hydrologists or people in radiative transfer, they all interested, or cloud physicists, and they're all interested in participating this improving parameterizations. And so starting from middle 1970 there is a big emphasis in improving parameterization and World Climate Research Program was organized so that we make some kind of field observations and satellite observations.

Edwards:

When did that happen? Do you remember?

Manabe:

And this is one, I brought it in here, which is this joint organizing committee physical basis of the climate and this is, so this here they recommend blueprint for improving parameterization, and this is exactly middle '70. Okay? And they discussed what is the most important problem, because they say we hardly know about clouds and we hardly know about cumulus convection. We have to improve it. So that is this thing, and if you look at this you can see within the variability, design of climate model, and then they talk of modeling and so forth. Then there are various processes they discuss. And so that the basic blueprint for improving physics parameterization of the various processes are laid out here. Because they look at my model, and obviously what I was proud of was an essentially the thing they say hey, we have to do more about this. What Syuki did was oversimplification. Right? And so that's how it started, and Arakawa was one of the persons who preached that how important cloud process is. So [???] Arakawa here, right here, here is an article, and he is saying how important the stratus [?] cloud is in addition to cumulus cloud. So he has a thing. But then you see if you look at my article which is in this same report, you can see I was ready to plunge into new venture of using this model for applying this model which I already have in —

Edwards:

Yeah. Well, let me just stop for one second and say something for the tape so that it is clear what publication we are talking about. This is called The Physical Basis of Climate and Climate Modeling. It's GARP Publications Series No. 16, published in 1975. GARP is the Global Atmospheric Research Program.

Manabe:

Okay. I also put a mark in here for your convenience, and I put [???] here. Oh, this is too late already if you look at here. Right here. This is differences. Now, this is my article there. And you can see at that time I already laid out instead of improving parameterization I was going to use this model, and my plan of how I am going to use this model, and how, what I have done is all spelled out so my second half of career is everything spelled out there.

Edwards:

This is what you were talking about yesterday, the place where you sort of lay out a plan and [???].

Manabe:

That's right. And first I show how good my simulation is here. If you at here, the simulation is very impressive.

Edwards:

This is Professor Manabe's article, "The Use of Comprehensive General Circulation Modeling for Studies of the Climate and Climate Variation" in the publication I just described.

Manabe:

Yeah. So you look at this, and this is [???] classification which I simulated. But then I began to talk about newly created coupled model. And I said how and when to use this coupled model for the study of anthropogenic changing climate and or as well as natural climate variability. I spell out all these things in the middle '70s. And so that was my past. And so that my seemingly sort of my attitude which looks like I am sort of indifferent to all the effort other people is making in parameterization is because of this situation I was in, because I can't do both. If I do improvement of parameterization, I will not be able to do the other one. And what I thought at that time is for the reason I spelled out already, that parameterization is important. However it's not easy and —

Edwards:

Now, in the period, in the mid-'70s, did anybody pick up the GFDL model and begin to improve its parameterizations somewhere else or here?

Manabe:

Well they didn't quite, I guess — I wonder who actually. We didn't have a policy to share our model with other people at that time.

Edwards:

Uh-huh. So does that mean you had a policy not to share it?

Manabe:

We are not quite ready to share [with] other people, because there was all kinds of problems, and even though the simulation looks good. And furthermore, the model which we had cost an enormous amount of computer time. You have to remember at that time nobody else, no other institution had computer to use our model as we constructed it.

Edwards:

Because this was running on a UNIVAC 1108? Is that the [???]?

Manabe:

Yeah. We are running day and night.

Edwards:

Yeah, and other people were mostly using CVC machines and Crays.

Manabe:

No, not even that. You know around the time, early 1970 for example, the other people hardly has much.

Edwards:

What about NCAR? Didn't NCAR have CVC machines by then?

Manabe:

Yeah, CVC 6600 and NCAR may be running too, but NCAR had their model by then of course. Although they didn't do the systematic simulation like we do. If you go through the literature, you can see that our simulation looks great and well documented. NCAR's simulations are rather, papers are rather fragmentary following our work usually, and simulation didn't look that good. And but anyway they developed their own model. And so the people who had a good enough computer, like NCAR they developed their own model, UCLA of course. They didn't have a computer, but Yura [?] Mintz, with his ingenuity he managed to get one computer after another, and so that he was able to run the model, their model, on the computers which were available.

Edwards:

What about the NASA groups?

Manabe:

NASA group was started from middle '70, right about that time, and Joe and Yura Mintz, you can ask Joe about it. Yura Mintz discussed — no. Joe and NASA, Jastrow [?], discussed it, but at the end they decided to use UCLA's model at GISP. So it's not much matter of a sort of giving model available to large number of users, which is the present situation. At that time it was not many people could use it, and also if you wanted to apply this model for some problem, it's very time consuming. So I had to do all kinds of idealization in this fast GCM study of Greenhouse Warming which published around 1970. And at that time everybody said is how stupid I was in using GCM, which is extremely time consuming. Everybody thought they should use much simpler model. So at that time I was criticized for using too complicated a model, okay? Which is a completely opposite situation from what it is now.

Edwards:

Yes. Yeah.

Manabe:

But the reason why it happened is at that time I myself found out to what degree model was successful in simulating the behavior of general circulation of the atmosphere and the climate and therefore I was convinced the time was right to use it for this purpose. However, nobody else believed in it. And so everybody else thought that I was in a kind of deep end "because he has a fast computer, he was spoiled by hardware and doing this mindless experiment." And everybody thought I should be doing much more intellectually oriented simple model study.

Edwards:

Do you remember particular people who said that kind of thing [???]?

Manabe:

That didn't say in that critical tone, but for example our friend Steve Schneider is one of the advocates of a simpler model, and he was advocating. So he was a very diplomatic guy, so he always say, "You have to use hierarchy of model, the simpler model as well as complicated model as well as simple model." But most people thought that it was premature to use GCM for such a purpose, because they thought, "You've got to improve parameterization first." Okay. So now then, some 20-some years later, I stayed with my old parameterizations, okay, and did all kinds of its application during that time when most people were working on parameterization improvement. Right? Now gradually the '90s come along, or maybe the latter part of the '80s, people started using GCM and now it's getting full bloom. But I must say that therefore then people suddenly discovered here is a guy who is sort of using his old very simplistic parameterizations still, okay, and seems to get very interesting results. They admit actually. But then —

Edwards:

[???] person who has followed this strategy more recently is Jim Hanson.

Manabe:

Yeah. Jim Hanson —

Edwards:

What people often say about him and the GISP model is that it is simpler and uses a larger grid scale than many other GCMs, but it's very good.

Manabe:

It's not as simple. But what enabled him to use, what it's characterized [by] is a large grid size. What enabled him to do that large grid size model and still get some decent results is because he used Arakawa's finite difference scheme. And not only the Arakawa. Arakawa proposed not only the energy conserving system, but also he proposed what they called square boticity [?] conserving, or potential instrophy [?] conserving system.

Edwards:

Okay. Can you explain those two things?

Manabe:

The energy conserving is to try to have a integral constraint in his finite difference scheme so that the earlier integrated the energy is conserved by through when you calculate advection [?] time. That means you calculate change of any kind of variable due to the advection by air flows. So this is what some people call nonlinear terms.

Edwards:

Okay.

Manabe:

And when you integrate all the area, kinetic energy conserves [?], that means your finite difference is crafted in such a way that velocity doesn't jump up. Because if one grid point bursts [?] and jumps up, then you square them, kinetic energy becomes enormous. So that Arakawa was ingenious enough to create the scheme in which this kinetic energy, suddenly one grid point shoots up exponentially by some kind of nonlinear interaction of hydrodynamical equation by doing this energy conserving system. Now however he further developed this system in which square boticity conserves. That the sheer [?], boticity is like sheer like thing, and this, so sometimes you get a thing called noodling that you get a very irregular flow which is very patchy and —

Edwards:

Noodling?

Manabe:

Yeah. Noodling. Yeah. Very irregular flow. And when you get too much of them, finite difference cannot handle hydrodynamical equation computations. And by square conserving, he eliminated this local noodling, extreme noodling. The finite difference is crafted in such a way to avoid this thing. And so Jim Hanson has a benefit of using finite difference system, which [was] developed by Arakawa, and therefore he gets away with rather coarse resolution models. Now he doesn’t share as much, or his group doesn't share as much about as toward the simplicity as I do. So his physical parameterization is pretty complicated. However he himself is in some sense shared very similar philosophy with mine, because very goal oriented. He knows what he wanted, and he don't care to use the shortcut. One of his short cuts is cross resolution model. And so he had some pragmatic approach. If he want to show his science, he can take any shortcut he can get. So he couldn't get the kind of computer Joe Smagorinsky managed to get. So that's why he had to do cross resolution model. And if his group of people want to make model complicated, he let them do it anyway. But he himself has agenda. And so that is I think more or less, my observation of Jim Hanson's behavior, explanations. And so that's sort of more or less explain my attitude in parameterization. But basic thing is also I believe in parameterization improvement also. Just because I didn't do it doesn't mean I don't believe in it, right? You can't do everything. You have to choose what you do. However, I do feel that I don't believe in the developing very complicated parameterization beyond your understanding of the phenomenon. That is, if you don't know enough, don't put it in. And that is my philosophy. Other people, even though you don't know enough, if you think it may be important you better put [it] in. I don't believe in that. Because you put unnecessary complexity, you get too many parameters in the model, which you have no hope of tuning. Right? And people say oh no, you don't have to tune. You can physically determine the magnitude of the parameters. That's a lie. And so you have no hope of tuning these parameters. And when you get a result, you don't understand why you get the result you got. Okay. And so that basically what I don't believe in is cluttered with the parameterization of physics, which you poorly understood. And all it achieves is throw down the computer, time to get a result gets much longer, and when you get the result it's incomprehendable [sic]. And so what I say to those people who make a premature incorporation of complicated parameterizations that if they can parameterize, you know, then the result projection which they get about Greenhouse Warming or whatever was not accompanied by understanding of the result which you get. And I say that kind of projection is indistinguishable from the projection by fortune teller.

Edwards:

Let me take a quick — [tape turned off, then back on...]

Manabe:

Yeah. What I think is that improving parameterization is a very important issue, and I think that we have to do everything we can to do that. But on the other hand that model is a little bit like stereo set, and you don't spend disproportionate time on the complicated parameterization of some process which you don't fully understand. And therefore —

Edwards:

How is that like a stereo set?

Manabe:

The reason why I'm saying this is, in the stereo set if you put a hundred dollar speaker on one thousand dollar amplifier.

Edwards:

Right.

Manabe:

And the people, when they construct a model, people usually put these things and the parameterizations they are working on, they have tunnel vision. Most important thing in the world. Okay. So that they want to put it in full complexity.

Edwards:

Right.

Manabe:

Okay. When there is other process which is equally important are treated very poorly or very simply. Now why you want to put the, you know, when you have a hundred dollar speaker, no point putting ten thousand dollar amplifier. Correct?

Edwards:

Yes.

Manabe:

And this kind of sort of balancing, what is important in the climate modeling is one of the important things is a balance. You just can't create a climate model which doesn't have a balance there. The same way as a stereo set. So if we are ignorant about something, you shouldn't put disproportionate complexity just because your parameterizations are complicated. And this kind of idea of balancing, people never seem to have cares about, the people who make a model. And the committee who make a model in choosing parameterization. Then they say okay, we are going to put most sophisticated parameterization in the world in this thing. And so as a result balance is not the most important thing. They put every possible complexity they can put in because on the paper these parameterizations which are all seemingly important process in, is the most advanced, best sophisticated parameterization — which is opposite from sophistications. And so they put more and more in, creating extremely complicated model which no graduate students or no scientists with ordinary means of computer power can use it. And the question then is the organization which is developed as a focal point of a university collaboration want to make such a model is the question I would say.

Edwards:

Yeah. So you're thinking primarily of NCAR when you say this or —?

Manabe:

Maybe I would say that, yes.

Edwards:

Are there other modeling groups that you think that [???]?

Manabe:

Other modeling groups may be different mission. But with regard to NCAR, this is I think is the NCAR is developed through NCAR UCAR arrangement — University Corporation of Atmospheric Research. National Center for Atmospheric Research funded by National Science Foundation. And so they should be able to provide what Steve Schneider called hierarchy of model with different degree of complexity. And so that university scientists with different problems at hand can pick whatever the model they can, they feel suitable as a tool for the problem at hand and be able, with minimum modification, they are able to use it. And that's what they should be doing; put more emphasis rather than developing all-encompassing extremely comprehensive model on the super duper computer. And that is related to — So my criticism of their things related to my philosophy, problem solving and the parameterization.

Edwards:

Let me ask you a couple of questions that are related to this, because you know what you are saying are that you began to be interested in using the model for things.

Manabe:

Yeah.

Edwards:

And I wanted to ask you about this paper, which is called "Climate Modification and a Mathematical Model of Atmospheric Circulation." This is a paper you wrote with Leith Holloway [?] in I think 1974, somewhere around there?

Manabe:

Yeah, yeah.

Edwards:

And the introduction to this article talks about climate modification rather extensively. It says, "In view of the far reaching social and economic consequences of climate modification, one should not attempt to modify the climate unless he can predict exactly the results of such an attempt."

Manabe:

Yeah.

Edwards:

This made me curious about the motivation for this piece and others you were working on around this time. You know, we talked yesterday about the study of critical environmental problems and the study of man's impact on climate which were both, you know, interested in the issue of human effects on the global climate.

Manabe:

Yeah.

Edwards:

You know this phrase "climate modification," there was a lot of interest in weather modification in the '50s and '60s.

Manabe:

I think that is right.

Edwards:

And is that where this came from? Or is this more related to the Greenhouse Warming?

Manabe:

I think this was the conference which was organized, all kinds of modification, I think including weather modification at that time. There were people talking about all kinds of modifications. And so I thought that I will — Usually I try very hard to make my talk consistent with the objective of the conference. So that I said, "Okay, I will think about one day and try to make it as relevant as possible to the theme of conferences." And yeah, so what did I talk about? Is this inadvertent modification here?

Edwards:

No. This is, this would be "Proceeding for the Joint Symposium, Centennial of the U.S. Weather Service."

Manabe:

See? So there were a lot of weather topics in there. And so I was invited as an add-on to this modification conference from climate viewpoint, so I concocted this paper. So this is not one of the papers I am very proud of.

Edwards:

[laughs] Okay.

Manabe:

But the weather modification is one of the beautiful examples what we shouldn't do. And because what I meant is I'm not saying weather modification we shouldn't, but there was a big controversy how effective the weather modification is. And they over-promised the benefits of weather modifications and later get a big debate about what they are doing is statistically significant or not. Finally they almost lost complete confidence by other people in the activity of weather modifications, and the entire field, not entire but the field of cloud physics which we need very badly now in the cloud parameterization, cloud microphysics I said hardly we know not enough. We don't know why the raindrops fall on our head; therefore we need every cloud physicist, good cloud physicists we can get in order to improve the cumulus parameterizations. However, the cloud physicists gave up because society suddenly became cool towards the idea of weather modification. And this was probably last, toward the end of that era.

Edwards:

Right. Yeah. My impression is that all the government programs on weather modification were dropped around 1975.

Manabe:

That's it.

Edwards:

Because they tried it a lot during the Vietnam War and it didn't, they couldn't [???] could have done anything.

Manabe:

Didn't work. Yeah. It's very difficult. That is a wrong story. I can cover forever on this one. But it's, even though I'm not a cloud physicist. But it is relevant to discussion here is later what we say to politicians.

Edwards:

Yeah, yeah. Well were people talking about climate modification in this period? Was that something that came up?

Manabe:

Not much. Not much. So, I was a little bit of a far-fetched outcast in this conference, obviously.

Edwards:

Okay. Yeah. But the Greenhouse Effect, you know, starting with Skep [?] and Smick [?] and maybe a little bit before, people had begun to think about the Greenhouse Effect. Oh and this, actually this reminds me —

Manabe:

Yeah. But you remember —

Edwards:

Oh, two things. One is the guy you were talking about yesterday named Carroll [?]. That was Carroll Wilson.

Manabe:

Wilson. Yeah. That's right.

Manabe:

At MIT, who organized Skep and Smick. And he was a member of the Club of Rome.

Manabe:

Mm-hmm [affirmative].

Edwards:

And one thing I wanted to ask you about from this period, early 1970s, were the limits to growth studies. Because there you know are some very similar issues to the kinds of ones you've just been discussing about modeling. And they had very comprehensive models. They tried to include a lot of things in their models, many of which they didn't know much about, because there was very little data on things like population and agricultural productivity and natural resources and pollution and so on.

Manabe:

Yeah, yeah, yeah. Yeah.

Edwards:

Was that something — were those studies things you were aware of and [???]?

Manabe:

Yeah, I was aware of, and I thought that that was one example where I shouldn't be getting into. That's what I thought at that time. Because it's very idealistic. Because you can put everything, not only the physical world but also social science regime, they want to put into modeling. And there was the Kennedy Administrations and with Kidd [?] and so forth, McNamara, everything depend upon the [???] calculates in the thing, you know, and this kind of thing, but you know as I watch this, many of the projections all seem to be the catastrophe and become ex — Of course there is a logistics equation, so you know everything, anything exponentially growing and we're better off. But when they have so many exponential growth in there, by the time it's leveled off it's bad enough. Right? So that by its gradually looking more and more like Don Quixote to me. This has nothing to do with Carroll Wilson though.

Edwards:

[laughs]

Manabe:

Carroll Wilson was a marvelous manager of this conference, two conferences, one at MIT and Smick. He did a marvelous job of getting people together. But in terms of — So I'm separating this criticism from his things, but it seemed to me that human being has always built in negative feedback. They are not as stupid as these models would indicate. Right? Now if you apply this thing to bacteria — right?— growth of bacteria inside bodies, or maybe even that may be difficult, but you know some simple, primitive living creatures, and there is a much number of them, and then how they grow and decay with time, it may be possible to do something somewhere, but it's [???] biologist may be interesting. People like [???] or along this line.

Edwards:

Right.

Manabe:

But to apply this to human things. And furthermore when they started talking about how consumption of coals are going to be and so forth, and then come about middle of 1970 and you know Energy Crisis, all of a sudden bmf! you know, the consumption of energy suddenly started leveling off. Nothing like. And what more which model with [???] can predict this? And so that there was some hiatus and enthusiasm about this model around middle 1970, and also [???] coincidence. But atmosphere, since 1940 to 1975, Northern Hemisphere temperature cooling down. Even though Southern Hemisphere temperature keep going up. And this is what I mentioned yesterday, people like Steve just saying Ice Age may be coming. But that was a minimum point and then it started going up again. And some people think the reason why the coming down was that sulfate aerosols started coming out from background and started cooling the atmosphere. In addition to natural variability, which have many marked [???] fluctuations. And so that's —

Edwards:

Is that what you think?

Manabe:

Oh yeah. I have a paper here I can give you. But so that's what the, yeah. And so.

Edwards:

It occurred to me after we talked yesterday.

Manabe:

Yes.

Edwards:

One could make an argument your work is responsible for the early awareness of climate change as a political problem.

Manabe:

Right.

Edwards:

Because you were the one who wrote the chapter in Smick on general circulation modeling and you had begun to do this work, Greenhouse Warming and so on.

Manabe:

Yeah. Yeah. Yeah.

Edwards:

What is your memory of how that became a political issue?

Manabe:

This is the, it is sort of — Originally, as I told you, I never worried about sort of societal impact of Greenhouse Warming. I was doing it mainly because, yesterday I began to explain, that is Greenhouse gas is second most important factor in climate next to sun.

Edwards:

Right.

Manabe:

And therefore, and it is warming up this climate, and by 30 degrees Centigrade by its existence, and therefore habitability of this planet doesn't exist unless you have Greenhouse gases. And that's why I was studying it. So, the thing is, and then I said — Yamamoto and Fritz Mueller stimulated me into doing it. But I was doing it, you know, because this, I see all these interesting papers about man's impact on this thing until that point. I thought yeah, I'm stupid not to take this opportunity to do something, dip into this problem. So you know, after I did enough of this Manabe-Strickler thing we did, I got into this Manabe-Wetherold thing, and putting water vapor feedback in, and then tried to see, evaluate the effect of these things. And I thought at that time I was stupid not to take this beautiful tool I created, and might as well write this paper on this thing, knowing what has been published.

Edwards:

On human impact on climate.

Manabe:

Right.

Edwards:

Knowing what has been published? What had?

Manabe:

Well I knew by then of course the work starting from Callendar [?].

Edwards:

Right, yes.

Manabe:

I didn't know Arenius [?], the funny thing is. But then, then later I found out that Arenius' work and Callendar's work are completely [???]. Arenius did some kind of radiative equilibrium. But then I went to the stockroom for Arenius' 100th Symposium and then scrutinized —

Edwards:

Yeah. That's not very long ago.

Manabe:

Yeah. Arenius paper. And found out that basically so far as surface balance perturbation analysis is concerned, this is no different from Callendar's and all these people. So these people are stimulated by Arenius until my, up to my point. And at that point then I put this, instigated by Fritz Mueller, and these difficulties encountered by the method which my predecessor has. So then I start putting effect of convection, how they change in response to increase of Greenhouse gases, not just radiative perturbations. So that's what — It's very little to do with my concern about social things. And I think the realization this may be social impact, that Carroll Wilson and all these people invited me [to] all these things, and after that the Department of Energy people like Dave Slade [?] and all these people —

Edwards:

Strait? S-t?

Manabe:

Slade. S-l-a-d-e.

Edwards:

Oh, it's Slade, okay.

Manabe:

Yeah. And these, at the Department of Energy organized many, many symposiums during 1970. They published many technical reports from Department of Energy. And so they, starting from around 1970 they have a continuous meeting or so. And so these people really stimulated me into thinking that this may be societal impact. But throughout this period, even now that I am not sort of crusading that sort of Greenhouse Effect is unilaterally harmful and therefore it should be avoided even though it may be very expensive, I am not sure of that even now. My personal feeling is, okay, now if CO2 really quadruples by business as usual, then climate change are so large I think approaches the days of Cretaceous period towards the end of dinosaur era. But even then, maybe human race can adapt to the environment. And you cannot probably, the countries like Canada and Russia may benefit from it. That's why my friend Budiko says. And furthermore because the CO2 concentration in the atmosphere is increasing and therefore in terms of photosynthesis if you come up with good biotechnology developing various agricultural species which can exploit this increasing atmospheric carbon dioxide for primary productivity then we may be able to utilize this increase in carbon dioxide itself. And furthermore if climate goes up and then growing season becomes longer. And right now we are close to Spring Equinox which is annual average sunshine. Even though we feel like we are in winter. Now if this were warmer, already we can be planting. I mean enough sunshine there and enough CO2 of course, more than enough. Maybe. But so that with considering how rapidly biotechnology is improving, but it's not obvious that we cannot adapt. But on the other hand, if you think that we have to have a biodiversity, which we have now, we would like to protect this, we would like to keep all the natural living things we have now, as many as possible, and we can't stand our ecosystem changing to something drastically different from what we have now, then if that is the value, then we have to — But the thing is that what I see now is the different countries have different agendas. Developing countries cannot afford to worry about global warming, they feel. Many. Like China and many people feels like this. Or some country like Soviet Russia may think, Russia may think, some people living there, "Oh, it's so cold in winter. If this were 10 degrees warmer."

Edwards:

That's what Arenius thought too. That would be nice if it were [???] degrees warmer here so we can —

Manabe:

Yeah. Warmer. So the point I am making, I seem to have emphasized all the benefits so far. However the shortcomings, the negative aspect of Greenhouse gases increase has been emphasized by a lot of people. And just balancing these things again. And so the bottom line here is that there is sort of benefit to some people. And this benefit well, negative aspect in other countries, and therefore it is all mixed. Okay? And so that if you try to get the international agreement, you can get international agreement so long as it doesn't cut into your bones, so long as you have your skin people can go along. However once it's getting more and more severe constraint in the way we produce you know industrial activities and all these things, then international agreement, it's getting more and more difficult — even though Greenhouse Warming gradually get out of natural variability. And so that I think is why it's so difficult to reach agreement. And what is worse is that people think that they know, people think that it is extremely difficult to predict future development of biotechnology. And you can see how rapidly biotechnology is developing and all others things, other kinds of technology, so that the thing is that people say it's difficult to predict future climate change, but it's even more difficult future change in our ability to adapt it to the change in our environment, okay? And therefore then in my opinion it is very difficult to get international agreement in a drastic action it takes. And therefore the important thing is we have to be open-minded about possibility of adaptations.

Edwards:

Going back into the mid-'70s, you did, you began to use these models and you began to do two kinds of things. You began to do more and more studies of effects of doubled CO2

Manabe:

Right, yeah.

Edwards:

And you also began to paleoclimate.

Manabe:

Yeah, yeah.

Edwards:

Tell me something about both of these. How did the doubled CO2 studies, what stimulated them, why did you keep working in that area, and what led you into paleoclimate.

Manabe:

The paleoclimate, my interest started — there is this meeting which was in Princeton. It was in somewhere middle 1970.

Edwards:

He's talking about the book Understanding Climatic Change: A Program for Action, National Academy of Sciences, published 1975. Okay.

Manabe:

Yeah. And it is co-edited by Yura Mintz, later Yura Mintz and Larry Gates. And many of the modelers, some of the modelers at that time, Larry Gates and myself and Yura Mintz, they all get together. Not that many at that time. And get together with Jim Hanson just studying around that time. So we all got together with some of the soft rock geologists who were interested in —

Edwards:

Soft rock? Is that what you said?

Manabe:

Yeah, the sedimentary rock geologists and or people who are interested in ice cores and so forth, they got together and they sort of inspired us about how big a change of climate occurred in the past. And on the other hand the modelers attended the meeting and told them how good model simulation is. So I showed one of these and by 1975 I had beautiful simulations, so I showed them you know, "Look here. Model can reproduce this." And so we had this very nice gathering in which sort of we stimulated each other and the result, one of the immediate results of this was John Imbrie —

Edwards:

Imbrie, I-m-b-r-i-e?

Manabe:

Yeah. John Imbrie, who was doing at that time he developed a method to determine the sea surface temperature of the [???] particularly maybe last sea surface temperatures during —

Edwards:

How? Is this —

Manabe:

By doing this he looked at there is a species of planktons. This is gradient plankton, not phyto-, oh it could be phytoplanktons, a species of plankton. The one he used most is fallumniphera [?], a special kind of they call it forninlomine [?]. F-o-r-n-i minefera, l-a, l-a or something like that.

Edwards:

F-o-r-m-o-i-n-f-o-m-i-n-e-m-r-a [?; I couldn't follow this—sorry!—Transcriber].

Manabe:

Yeah. They call it fornim [?] usually for short. And then the fornim has a different species, and they seem to be most sensitive to sea surface temperature. They swing up and down in the, I guess, eating these phytoplanktons. But they are very sensitive to the temperature of the environmental water, surrounding water. Okay. And so different — by looking at taxonomic [?] composition of fornim and then carry out mathematical analysis of that taxonomic composition. Some kind of factor analysis. Or sometimes they call [it] [???] factor analysis. They are able to reconstruct the temperature during Ice Age for example. And so Imbrie shortly after this meeting persuaded a large number of geologists. He is a sort of charismatic guy who is capable of persuading people to work together to try to create mapping of sea surface temperature at last glacial maximum [?]. And then other thing he did was to look at the pollen distribution on land surface, what is sort of climate over continent must be. And then you get people who does, a glaciologist [?] who can determine the margin of continental ice sheet during last glacial maximum and so that he can use these paleoclimatic reconstructions of last glacial maximum for evaluation of —

Manabe:

...work that John tried to sort of, uh, that many of us are very much enthused about, mainly because the climate of last glacial maximum is so different from what we have now, and it really, this is a sort of good testing ground of our models. And so in a sense the talk I gave them, and other people's talks, was successful in convincing paleoclimatic community to go into this. Because usually paleoclimatologists look at one cause or something, look at one cord [?] rather than mapping.

Edwards:

Oh, really?

Manabe:

Yeah.

Edwards:

So there was an existing field of paleoclimate, but it tended to focus on local areas or regional areas.

Manabe:

Yeah, yeah, region. Yeah.

Edwards:

I see. Or pollen distribution that —

Manabe:

They have been doing all of that. But in global mapping.

Edwards:

Very interesting.

Manabe:

This is a sort of beginning of new eras in that kind of thing. And later John Kuchback [?] who is coming to this meeting starts talking of co-map [?] experiment so that since last glacial maximum he has a time slice of various time points every so many thousand years and then try to reconstruct these climates.

Edwards:

Co-map? What does that stand for?

Manabe:

I wonder. Co-map. Some kind of a Holocene [?] mapping of climate.

Edwards:

Okay. It's h-o-m-a-t?

Manabe:

Yeah, [???].

Edwards:

Oh, homat. Okay.

Manabe:

Holocene. No, co-map he call it, so — yeah, so I don't know what first C is, but — So anyway, so this is a kind of a new era which is sort of inaugurated by, started from this meeting, and the time becoming gradually ripe. And John Imbrie wanted to collaborate with me immediately after this, but I was sort of typical sort of Japanese sort of understatement, so didn't sound too enthusiastic, although I was actually enthusiastic, but I wasn't too enthusiastic, "Oh, I have to do this first and then I can go to that," and John couldn't wait for that. So he went to Larry Gates and Yura Mintz and used 2-level models of the Mintz-Arakawa model to —

Edwards:

And Gates was then at Rand or had he gone somewhere else?

Manabe:

Yeah. Rand Corporation, yeah. So that's what he started doing it, and Gates did given this sea surface temperature.

Edwards:

Did what? I'm sorry.

Manabe:

Given this sea surface temperature. He tried to determine how the response of climate model over land surface is. And then compare with pollen later. And I did a similar calculation, slightly later but about the same time. But climate, official calculation was Larry Gate's calculation. But then later I started using this sea surface temperature as validation data for my output, so that instead of giving sea surface temperature I must try to get land temperature. But using atmosphere coupled with a rather simple model, I try to simulate this climate-determined sea surface temperature.

Edwards:

Okay. So you are working in the other direction.

Manabe:

Yeah. And that I did pretty good work on that. I have published several papers with Tony Broccoli. His name is the same as broccoli, it's a vegetable.

Edwards:

Yes.

Manabe:

And then try to understand why temperature, what factor contributed [to] this cold climate of Ice Age.

Edwards:

So you were working from the gaseous composition of the atmosphere?

Manabe:

Yeah. Gave the Greenhouse gases which is lower, and then put a —

Edwards:

Okay. And that data comes from —?

Manabe:

Ice core they come from.

Edwards:

Ice cores.

Manabe:

Yeah. From measurement of ice cores. There is [???] bubbles.

Edwards:

In the ice cores.

Manabe:

Yeah. And so I put that in, and then I put the continental glacier, which was determined by Joe Denton [?] and collaborators as a part of climate project.

Edwards:

Right.

Manabe:

And then I tried to see when you put the highly reflective shiny ice sheet and reduce the CO2 and other things can I reproduce the temperature, sea surface temperature, which John Imbrie and climate group produced. And so then I started doing strip show approach. Take one factor at a time; put it in or out and out and so try to understand the cold SDD or land surface temperature. Much colder at that time during last glacial maximum. And then try to evaluate also the sensitivity of climate model by using these large climate changes. Can we understand this by using climate models? So that's what we are doing now, doing at the — After climate fever subsided, I keep pushing, pushing and for a while most people's attention turned away.

Edwards:

Really.

Manabe:

Yeah.

Edwards:

When? When did that happen?

Manabe:

Oh, shortly after. Once you know Larry Gates and the climate people published this monumental paper in Science.

Edwards:

Which was when?

Manabe:

Oh, around 1976 or 8, maybe 1978 or somewhere thereabout, about simulating land temperature comparing with pollen. I told you the area, given the sea surface temperature of climate rather than using sea surface temperature as a verification, which I did later. Then I sort of, emotion subsided and the climate group also changed into SPECMAP. They want to look at trying to understand rather than reconstructing last glacial maximum environment, they started seeing spectra, time fluctuation of end of many, many points which they analyzed through CLIMAP, but now they went back; at each point they look at the time series of these temperature changes at all these points.

Edwards:

Okay. So they're trying to, rather than just a snapshot of climate at a given time, they are trying to get the time evolution [?] of —

Manabe:

Time series. Yeah. And then sort of measure discovery follows that indeed Milankovich [?], astronomical theory of Milankovich, the Ice Age is indeed pace makers of Ice Age is the fluctuation of orbital parameter of this earth around the sun. And this is so the most shiny monument of John Imbrie's and other people's work. And so they went that way, and so that for a while sort of detached from modelers. And —

Edwards:

Because they could do this without a model. They just needed to establish what the climate was like in the periods and relate it to fluctuations in the orbital parameter [?].

Manabe:

Yeah. That's right. Yeah. And much later, people like John Kuchback began to sort of look at this SPECMAP data and then try to —

Edwards:

Spell that?

Manabe:

Kuchback. K-

Edwards:

No, no, Kuchback's name.

Manabe:

SPECMAP, S-P-E-C-M-A-P.

Edwards:

Okay. Mm-hmm.

Manabe:

And so forth. So their attention turned to SPECMAP, which is a great effort there, which was very fruitful, great meaning fruitful. Now meanwhile however, I started digging in more and more in collaboration with Tony Broccoli try to use these great data of CLIMAP for a wide variety of modeling study. And this, I'm very satisfied with this. It was very useful that all kinds of interesting insight about why the climate of last glacial maximum was the way it was. And also after a while then as 1990 comes in other people, there are by now many, many modeling groups all over the world, and they begin to [get] interested in the value of these climate data, carrying out similar experiments as Broccoli and I was doing, as well as original experiments which Larry Gates pioneered. And they tried to do this what they call PMIP, Paleoclimatic Model Intercomparison Project.

Edwards:

When was that?

Manabe:

That's started from early 1990. So International Paleoclimatic Model Intercomparison Project. So that is what sort of history of CLIMAP and the modeling. And more recently of course —

Edwards:

And where was PMIP done?

Manabe:

PMIP is being done internationally now as a part or — I don't know, probably as a part of PAGE [?], Paleoclimatic Analysis and I don't know what called there, the PAGE, an international program. You ask Steve what PAGE stands for. Okay. And so as a part of PAGE, and PAGE is a part of IGBP, International Geosphere Biosphere Program. And also PMIP may be a part of CLIVA [?], International Climate Variability, modeling experiment. I am not sure exactly how it is done. All these collaborations are international. But anyway, the PMIP program comes out, and so that was how this development following this meeting which occurred in the middle of 1970 is at present.

Edwards:

Okay.

Manabe:

And of course I got more interested in, more recently I have been collaborating with Wally Broecker, who is coming here. And —

Edwards:

Spell the last name, please?

Manabe:

Wally, Wallace, W-a-l-l-a-c-e B-r-o-e-c-k-e-r. He is an outstanding creative scientist. Both Imbrie and Wally is a sort of certain kind of charisma which sort of stimulates other people until they keep asking fascinating, challenging scientific questions and manage to get people motivated continuously.

Edwards:

Okay.

Manabe:

And so Wally is asking [about the] abrupt climate change which appeared in the ice core, isotopic temperature, and many other evidences.

Edwards:

Right, right.

Manabe:

And he thinks that this is related to rapid change in the intensity of some heron [?] circulations, that is conveyor belt they call it, some heron and [???] overturning of ocean circulations in the Atlantic Oceans, and this conveyor belt is the reason why the Scandinavian climate is much more mild, much milder than the Alaskan climate, which is approximately the same latitude. And so if this some heron circulation slows down and change in the North Atlantic and the surrounding regions, they much very change. And they have — So Wally again galvanized whole group of paleos [?], oceanographers, and all these people as well as modelers. And I have done all kind of interesting experiments. That's some of my, explain some of my recent publications on the Young Adrias or abrupt climate change or multiple activity of ocean circulations and —

Edwards:

Right.

Manabe:

We have published many, many papers. And I think that we stimulated Wally's and these people stimulating to many people. Oceanographers now getting into it, many of whom we are inviting so that morning of the first day you will find very interesting discussions of these abrupt climate changes. So that is another aspect of glacial climate paleoclimatic study. However, we are also interested in — Now the very warm climate, which is similar to Greenhouse-induced warm climate. And that's the time of dinosaurs. And why dinosaur climate was warmer at that time. And so these, something like 70 million years ago before the Great Extinction, which occurred about 65 million years ago. So why it was so warm?

Edwards:

It was about 5 degrees Centigrade warmer than now?

Manabe:

Yeah, it could be more.

Edwards:

Maybe more?

Manabe:

Much more. Yeah. Now, sea surface temperature may be warmer by that much, but the land temperature is much warmer.

Edwards:

How much?

Manabe:

Oh, in the polar regions, oh it will be like 7-8 degrees and global mean temperature, land temperatures may be — so that's 7 degrees global mean and maybe even more, and polar region temperature may be, Northern Hemisphere may be warmer by maybe 15-16 degrees Centigrade. Okay. And land temperature may be warmer by 10 degrees than now. But global mean may be 7 more. So. But I am just guessing you know, but it's a very warm climate and Butiko, the Russian scientist, one of them. And then more recently the guy named Bob Berner at Yale speculated that this —

Edwards:

B-e-r?

Manabe:

B-e-r-n-e-r. Professor at Yale University. Berner and Budiko and Ronov Aria [?] had this, some kind of a chemical cycle, bio too but geochemical cycle model which began to see what the key process which controls the atmospheric composition of carbon dioxide or oxygen or things like that over hundreds of millions [of] years time scale. Involved with silicate and other things. And some of them are weathering, and some of them are magnetism, which is called volcanic activity, which gets CO2 from the [???], and weathering are the ones fixing CO2 in the locks [?] then. And so got interested, he suggested the reason why during this cretaceous period, which is toward the end of this Mesozoic [?] dinosaur period may be due to the — the reason why it was warm was continental spreading rate was faster so that more magnetism, more out gassing of CO2. Okay. And the reason he also thinks, the reason why latest katana [?] [???] ations, we are in a glacial period, we adjust to [???] with the interglacial, which lasted a very short period. And most of this climate is very cold. Most of the katanic [?] period is Ice Ages. And so why this katana [?] is so cold as compared with cretaceous or the time of great extinction, which is 65 million years ago? And he thinks that continental spreading rate becomes slower so that out gassing of carbon dioxide becomes less, and furthermore you have a huge mountain building, such as Tibetan plateaus, Indian continent will crash into Eurasian continent and [???] went up, and therefore weathering, which fixes your atmospheric CO2 into rocks become faster. Because of mountain building. The Himalayas, the European Alps, and the Rockies also at that time. I think the most important one is the Himalayas. And that's actually the fixing of CO2 into rocks also. Not only the throw down of continental spreading rate, but increasing weathering due to the mountain building. That reduced CO2 which created a cold enough climate in which ice sheets can grow. And so first stimulated by Budiko's Ronov study, we begin to do —

Edwards:

Budiko's what study? I'm sorry.

Manabe:

Ronov, R-o-n-o-v.

Edwards:

Okay.

Manabe:

Two Russian scientists. But Budiko is a friend, Ronov I never met.

Edwards:

Oh, I'm sorry, I missed that. I thought you were saying something else. Spell it again?

Manabe:

Ronov. R-o-n-o-v. This is a geologist. Budiko is a kind of a pioneer in simple climate models. And among other things. So they suggested that first gave the impression this is a long time ago. Actually it is around late 1970s or early '80s, somewhere thereabout, 1970s probably. And then Berners’s work is late 1980s, so stimulated by Budiko Ronov's work, Bryan, Kirk Bryan and I started putting coupled ocean atmosphere model, simplified version, to try to study the circulation and climate at much higher level of CO2. And now more and more people believe this, more recently, this Berner and other people like Gallo [?] is another one. These people, Berner collaborated with a few other people, but Berner Etal [?] I guess, Etal can be very important. But they collaborated with [???]. So and now their chemical cycle model made more and more people believe that CO2 may be indeed much higher during the time of dinosaurs. Some. And so, that way for it is going on to study these high CO2 concentration [???]. And again this is interacting very closely with trying to project future climate change.

Edwards:

How?

Manabe:

And so giving sort of inspirations that earth's climate was that high. And there is some paleoclimatic evidences like paleozoles [?] or the number of stimata on the leaves, stimata density on the leaves, which seem to suggest that CO2 concentration was much, much higher at that time. And so that [it's] nice to — it may not be nice, but it's sort of make people think that very Greenhouse rich warm climate really occurred in the past when we study these things there.

Edwards:

It has rarely occurred?

Manabe:

Really. Actually occurred in the past, has occurred. And again leaving us a marvelous playground using the model, evaluate the performance of model there. And so these kinds of studies are going to be more and more, going to be carried out much more. And what is happening now is that both glacial climate and climate of distant past like dinosaur age, there is particularly I think glacial climate, there is more and more sort of tool of isotropic analysis of the deep sea cores and analysis of chemical and isotropic analysis of ice core and they can date better the pollen analysis or they can date better deep sea cores, and then in addition there is a more distant past they use coral, coral [???], c-o-r-a-l.

Edwards:

Coral [???].

Manabe:

Yeah. And again chemical isotropic analysis of coral data, and so that with the variability of accelerator, mass spectrometer, we are getting much, much powerful tools to decipher how the past climate changed. And much more quantitative data about geological past. And therefore paleoclimate are going to be very interesting playground for all climate modelers. That's what I think.

Edwards:

Yeah, yeah. Well, let me take another quick bathroom break, and I have a few more questions about this particular thing.

Manabe:

Yeah. Right. Yeah.

Edwards:

Okay, the thing you were just saying about data on paleoclimate makes me remember another question I wanted to ask, which is more connected with present climate, and that is that during the period we're just talking about, the 1970s, the first satellite data started to become available.

Manabe:

Yeah.

Edwards:

And I'm actually curious about data sources for all your work from the '60s onward, but I'm especially interested in sort of how the availability of new data may have affected what you did.

Manabe:

Right. I [was] always sort of inspired by all these satellite data, but for these studies I hardly used them. Now of course some of the simulations of tropical general circulations by looking at where the ITCG [?] is in the satellite.

Edwards:

The what?

Manabe:

Where intertropical convergence zone, tropical rain belt is, and you can look at these global pictures of geostationary satellites. And then at least semi-quantitative/qualitative sense, we really felt that our simulation of global perturbation distributions are really pretty good actually. But I think the satellite data—but I really don't actually get into, get actual satellite data and the processing to try to compare with this, our climate model. Except a few exceptional cases in which like cloud forcing of satellite. One of the biggest uncertainties in the climate model is still cloudiness and its radiative effect. And one of the things was proposed by Ramanason [?], the concept of cloud forcing, how cloud is thermally, radiatively forcing the atmosphere. And these quantities they calculated, and this is one of the very effective ways to evaluate how model cloud simulations and its radiative effect was doing. And I think we just began to use this satellite data beyond these original inspirational uses. But begin to use it now, and but from now on other people try to evaluate the cloud parameterization and its radiative effect and all these things. They got to use more and more, increasingly more and more in connection with the treatment of these cloud radiation feedbacks and all these things are going to be very, very important problems. So satellite use are going to be increased tremendously.

Edwards:

Starting in the mid-'60s when you began to publish the results from GCMs, you often have a comparison between the computed and observed. Where did those data come from?

Manabe:

They are usually, most of these data taken in connection with weather observations and so that you have a rain gauge measurement in terms of rainfall, and then in terms of surface air temperature you have day-to-day measurement of surface temperature so that there is a nice computation done.

Edwards:

Alright. But by whom?

Manabe:

For example at GFDL. Yeah, GFDL there is a Bram [?] Oort, O-o-r-t, and Rasmusin [?] compared beautiful data. The beautiful data of the tropospheric statistics of troposphere, how tropospheric temperature changes with season and they also have how the momentum on the heat and water vapor are transported in the atmosphere. And they have a painstaking massive compilation of these observational data, which originally started at MIT under the guidance of Professor Starr.

Edwards:

S-t-a-r?

Manabe:

R. S-t, 2 r's. Yeah. Victor Starr.

Edwards:

Oh yeah, [???].

Manabe:

Yeah. And at the beginning there was the UCLA, Yura Mintz, and Yura Mintz was doing it too. But I think Professor Starr at the MIT group just kept on going on and on and on, and then sort of did a very comprehensive analysis. And then Bramold [?], who is a graduate student of Professor Starr came to GFDL and under the instigation and support of Bob White and Joe Smagorinsky he continued these massive compilations which sort of culminated into an atmospheric linear statistics book published from GFDL.

Edwards:

Okay.

Manabe:

And a professional report, [???] professional report or something. So that's one thing. And then there was some kind of young, obscure scientist, the guy named Sidney Levitus.

Edwards:

Spell that. Sidney —?

Manabe:

Levitus, L-e-v-i-t-u-s and Sidney.

Edwards:

L-e-v?

Manabe:

Yeah. Levitus, L-e-v-i-t-u-s. And the first name is Sidney. Okay. He was a staff of the Bramold's group, but he started compiling the statistics of ocean, oceanic temperature and salinity. And usually oceanographers also tended to look at local, interested in local data. So you send a [???], you get a beautiful, publish a beautiful atlas of that cross section where a ship travels, or a thing like that. But they seldom publish global compilations of the data. Afraid of some of the massive number of data available may be incorrect. They feel they can only trust very carefully measured selected cross section measurement can be trustworthy. So they didn't do this compilation effort such as the one massive compilation such as Sidney Levitus was waiting to do. So ordinarily they are very critical, but it turned out Sid creates a global mapping of salinity and temperatures, three dimensional mapping, and every season, yeah, near the surface he may have a different season, but oh more like annual mapping, annual mean mapping. So three dimensional mapping oceanic state. And put that in tape memories and it was one of the citation readers now. And this was a great thing for our coupled modeling effort, because [???] we needed to see how well our ocean simulation is. We can immediately go to his tapes and compare them very quickly. And it's helped us a great deal. So both Bramond's general circulation statistics, old statistics and Sidney Levitus had been the primary data source for evaluations.

Edwards:

Okay. And where it Levitus?

Manabe:

Levitus is now working on another branch of NOAA, environmental data service, data and information service of NOAA. They call it NESDEIS, N-E-S-D-E-I-S or something, Information Service. Satellite, S stands for Satellite and then Data Informational Service of NOAA.

Edwards:

But he was at GFDL for a while?

Manabe:

Yeah. When he was doing this compilation. Now he is doing sort of a deviation from that mean normally. And he is making much effort of compiling many, many data scattered all over the world, such as Imperial maybe, German maybe, all these things, Dustov [?], these, all the data, and digitized them and so that he can study the natural variability in the state of ocean which is of vital importance. We benefit a great deal from his data analysis. And so we can use these data for evaluating not only mean state but some of the interesting phenomenon which is produced by long mean time integration of coupled ocean atmosphere models. It is a great data. And so these are the major data sources. However what I am saying now is as people are getting more and more in the clouds or things like that and or a more recent sea surface temperature data, then satellite data we begin to use when we look at how sea surface temperature fluctuates over long time scales. We begin to use a sea surface temperature mapping which uses both conventional and satellite data, and so that is how satellite data is being used. But what I am saying from now, well, starting from more recent past but into future we are going to use satellite data for evaluating how model is simulating cloud and its radiative effect. And that is going to be extremely variable data, which satellites can provide us.

Edwards:

Okay. Let's see. We've been going for a couple of hours.

Manabe:

Yeah.

Edwards:

Let me see if there's anything I want to —

Manabe:

I think that what's left is some of the coupled model problems. And so coupled models, we talk about I get started, or we didn't talk about it, so we probably talk about coupled models like you ask questions about how we coupled them, what was the problem.

Edwards:

Right.

Manabe:

So we will go to that.

Edwards:

And then also I wanted to ask you about the IPCC.

Manabe:

Ah, so yeah. IPCC. And so then another interesting question is in connection with IPCC what we do in the future in order to reduce uncertainty, what we can do with model. And that we have to discuss.

Edwards:

Yeah. Okay. Well why don't we stop now —

Manabe:

Well these are the two things which are left I think now.

Edwards:

Right.

Manabe:

And then at the end you ask the question that what I think is the best contribution I have done, so that will go briefly. And that would be — right.

Edwards:

Okay.

Manabe:

And so there are three topics left I think.

Edwards:

Great. [tape turned off, then back on...]

Manabe:

...the model I think now. Right?

Edwards:

Okay. Yeah. Sure.

Manabe:

And the thing that I would like to explain first is that to use these models for —

Edwards:

Coupled models.

Manabe:

Yeah. Or any climate model to do project of future climate change or want to study the climate of distant past or anything. But first of all it is important to get realistic climate simulation of modern climate. Now usually when people say, meteorologists say they get nice simulation of climate by models, and they are usually talking about given the sea surface temperature to be observed. And then use that as a lower boundary condition and run the atmospheric general circulation model and get a climate which looks like real climate. And which I have been talking mostly until now, given the sea surface temperature, what kind of climate you can get, and even Larry Gates' original experiment with CLIMAP data, you give the sea surface temperature of last glacial maximum. But in order to really study how the climate responded to change in Greenhouse gases or change in orography [?] or change in sunshine or change in earth rotations, then it is necessary to study, first make sure that climate, including sea surface temperature, is simulated realistically. And then you want to perturb [?] that, and then also that then you can study the response of this model of climate, two various forcing, such as increase of Greenhouse gases. So that's what we have to do. And so then, so the goal of good coupled ocean atmosphere model is to get the realistic atmosphere as well as the ocean, including sea surface temperature. Starting from, usually we start from some kind of initial condition which is something like isothermal atmosphere, isothermal ocean at rest. So there is absolutely nothing predetermined. It's just a quiet atmosphere, quiet ocean, with a constant temperature.

Edwards:

Isothermal means it's all one temperature.

Manabe:

Yeah. Right. And then you turn on sunshine, okay? And then what happens is usually if you turn on sunshine you've got, tropical latitude becomes much warmer and polar latitude much colder. And as time goes on, so that this creates some kind of a density, no uniform density distribution of air.

Edwards:

Right.

Manabe:

Okay. So that then some air start rising, and some air start falling. And after time goes on since earth is rotating, so that it doesn't go simply warm air rises, cold air sinks, but it started twisting around because of [???] force, which you get as a result of earth's rotation. So what ideally you do is you start running this kind of coupled ocean atmosphere model which turns on sunshine, start from rather unbiased wind chill [?] condition. And after [???] the long, long time, first atmosphere temperature starts getting more uniform, and then gives that temperature, transmits that to ocean by heat exchange between atmosphere and ocean. Also winds start blowing in the atmosphere. That broad ocean surface water, wind driven ocean current starts emerging. So after you keep on going on and on and on, then you after a while you get the hot, warm tropics called the polar region, in the atmosphere you will have a jet stream, you will have cyclone waves, moves around, develops just like daily weather, and ocean you start having Gulf Stream across your current, all these ocean currents start coming out, and temperature of the ocean starts getting thermo climb, which is the upper warm layer, and the core of the deep oceans. All these things come out after you run a thousand years, several hundred years at least. But in order to get a nice equilibrium you have to run a thousand years. Then you get, if model were perfect, you get a perfect simulation of atmosphere and ocean and its general circulations. And once you get that, then you perturb with increase in Greenhouse gases gradually or suddenly and then see how this responds. And this is called a numerical experiment. And you do this numerical experiment for example when you want to project future climate change. So in order to get started on your numerical experiment, you first have to get a realistic system of coupled ocean atmosphere system to get started with this experiment.

Edwards:

Right.

Manabe:

And so the —

Edwards:

When did you first start to realize how important adding in this realistic ocean was going to be?

Manabe:

This is already we realized. When I was a graduate student I would study the Sea of Japan and heat budget of Sea of Japan. So I studied the atmospheric heat budget, ocean heat budget, make sure they are consistent, and so I didn't realize that ocean had an enormous heat capacity, and so that it takes a long time to adjust to its own equilibrium state and coupled with the atmosphere. And of course Kirk Bryan always preached to me how slow the ocean process is. Okay. So that —

Edwards:

When did he come here?

Manabe:

He came here in 1963 or something like that, maybe '62, somewhere around that time.

Edwards:

So while you were still in Washington?

Manabe:

Yeah.

Manabe:

...working together. But anyway. So, the problem here is that to run the atmosphere model with time, integrate atmosphere model with time, follow its evolutions. Takes a long time, because general circulation model has many grid points and you have to calculate everything. It takes a long time. So that it is very difficult to run atmosphere model of thousand years. So what Kirk and I worked very hard at the beginning is to, how to economize this process. You can't simply run — you can run ocean hundreds and hundreds of years. Even the computer at that time, if you don't increase the grid size, computational resolution too much, we are able to run ocean model reasonably well.

Edwards:

Because you use a much time step.

Manabe:

Time steps, that's correct. So —

Edwards:

How long were the time steps in this?

Manabe:

Oh, one day or half day or something like that. And also the areas of physical process. Ocean doesn't have a radiative transfer, and so many other things are simpler in the ocean than the atmosphere. And so then at that point we devised a method called, what we call a asynchronous coupling. This is a product of our collaborations, and I think both of us come up with many good ideas to develop this scheme. Both of us, both Kirk and I. And so when you run this thing, you run the atmosphere model one year. During that time, ocean model you run hundred years. And then you sort of, because the ocean is slow process, so you asynchronously couple the atmosphere in the ocean. And so without wasting, running atmosphere model a hundred years, when you run only [???] atmosphere one year, ocean you run already hundred years. And then by the time you run atmosphere several years, ocean is running several hundred years, and it reaches the equilibrium state of climate in the optimum length of time. And that is the most important innovation the paper I told you already 1969.

Edwards:

Yeah, I was looking at that and I was going to ask you what that meant because I didn't understand the —

Manabe:

Yeah. We come up with this optimum coupling. And also Kirk and I came with idea of how to speed up the deep ocean, because deep ocean takes much, much longer to approach to the equilibrium. And so that you would like to accelerate deep ocean even more. And we come up with all kind of ideas so that in a sense I said I didn't spend model improvement starting around 1965. That means essentially I didn't try to improve atmospheric model. When other people started improving atmospheric model I stopped that, used my simple, simple what I call simple well balanced construction of atmospheric model. I satisfied with it. But that doesn't mean we are lazy doing nothing in improving the climate model. We are spending all our ingenuity and time developing how to couple ocean GCM with atmosphere. And to get the realistic equilibrium so that we can carry out various interesting numerical experiments. And this asynchronous coupling may become even more important if you try to study why Ice Age, why huge ice sheet developed on this planet during last glacial period. And because when you talk about the atmosphere, the time scale of adjustment is several months. On the other hand ocean takes much, much longer, several hundred years. You have to run the coupled model to get coupled system equilibrate, and probably you have to run several thousand years in order to get deep ocean to equilibrate. But then ice takes also several thousand years. So ocean time constant maybe time scale of adjustment is actually several hundred years, but to get the perfect equilibrium I'd say several thousand years. So ocean time constant is several hundred years, but the ice sheet time constant is several thousand years, which is again ten times longer than the ocean. And so in order to study for example, evaluate astronomical theory of Ice Age, try to understand why we are now in the interglacial, 20,000 years ago we are the last glacial maximum. And before that there was a long cold period, although ice sheet grows and decays somewhat. But you have to go back another 100,000 years before this last glacial maximum which is 20,000 years. You have to go 100,000 years or more than 100,000 years before that in order to get last interglacial. At present we are interglacial, we have to go to last interglacial. Now, we would like to understand that, right? Why. And but in order to learn, there is no way even for the super duper computer. You can run the atmosphere model hundreds of thousands of years. So when you have a faster machine there is another way to use that machine capability. Instead of keep increasing resolution, everybody go to 10 kilometers. Some people could keep the resolutions the same. But to try to study the mystery of glacial, interglacial climate change is an Ice Age is coming as Steve Schneider once predicted? If so, how long into the future? What is the future fate of the Greenland Ice Sheet? What is the fact of Antarctic Ice Sheet? And what is the effect of Greenhouse Warming, how Greenhouse Warming would affect future probability of having Ice Age. All these questions have to be answered, although I am not sure whether human race will last that long the way they are behaving now.

Edwards:

[laughs]

Manabe:

So these are some of the interesting questions, and this technique or asynchronous coupling are going to be very, very important, even no matter how fast the machine is you always would like to have that kind of technique in order to get an optimized, economized computer time. And so that is, we are doing, starting from late 1960 all the way to, oh, latter half of 1980 we are trying to get the nice equilibrium. The problem is, even using these efficient techniques and get equilibrium, this equilibrium state of ocean is not quite as realistic as we want it to be. So we struggled about 15 years right there to get realistic equilibrium. So we are not lazy not improving parameterization as other people are doing. Arakawa was spending all his time, and many other people, some other people started doing, instigated by my Aria work on the land surface [???]. I am the first one to put the land surface process in. And they tried to improve on it. So they started working and working, and then here is, "Hey, Syuki Manabe," you know I may inspire them. But when they come up with these better schemes, he has no inclination to use the [???]. Because I was busy. Okay. And so that's what I was doing, and so then towards the latter half of 1980 then I realized how long, that what happened is that if these models, if you run the atmosphere model separately from ocean, giving the sea surface temperature, we can get a nice climate, as I explained to you. If you give the nice sea surface temperature and the surface salinity and run the ocean model, we can get nice structure for the salinity distribution of ocean simulated beautifully. Okay? But when you put it together, things doesn't produce as good simulation as we would like to. Because small area [?] will be magnified when you couple together, because more degree of freedom. So more way to go [???]. Okay. So I was struggling about 15 years. And then I say okay, now I waited long enough and my career may not last forever. Okay? And so then I said when some places surface temperature becomes too cold, I decided to systematically put a certain amount of heat in. When surface temperature becomes too warm, I put certain amount of [???] heating. And once you, this you can determine very systematically how much heat, positive or negative heat do you need and where.

Edwards:

This is flux adjustment.

Manabe:

Flux adjustment. You put that in. Then at least you get the realistic sea surface temperature, surface salinity distribution with sea ice about right place so that when you perturb you are perturbing around a realistic state. Ordinarily if you don't do it — See very often people, instead of starting from isothermal atmosphere ocean at [???], most people started from realistic [???] condition, put present value into the model and started from there. What usually happened is climate started drifting hopelessly away from realistic state. So that controlling the experiment became unrealistic, because then you are perturbing around unrealistic control experiment then you are not testing the sensitivity of the climate as we know it but some other climate. And this is a problem. And so that ideally I would like to have a play guard [?] and then say start from, this is a [???] experiment. We started from nothing, God put the sun in the sky, and we got this climate simulated. But there is here and there that some heron circulation isn't quite right, so Europe is not as warm as it should be as compared with Alaska say for example, Canadian coast, and so forth. There is a various problem between continent [?]. And so we did this flux adjustment around oh, toward the end of, about around toward the end of 1980. And then —

Edwards:

Now when you do the flux adjustment, is that based on comparison of the model behavior with realistic observational —?

Manabe:

Yeah. Now you started from, you first, the way we do is instead of the way I described it, the way you do then is you put a realistic sea surface temperature, and the land ocean model in the accelerated way, various techniques around ocean model.

Edwards:

You start to observe —?

Manabe:

Observe the sea surface temperature.

Edwards:

Okay.

Manabe:

As at the boundary condition [?]. Land ocean model. And so at the surface you have a perfect sea surface temperature distribution. And then subsurface ocean temperature is in equilibrium with the surface conditions. Now you run the atmosphere model with same sea surface temperature, which it observed, run it to equilibrium. Atmosphere has some circulation about it. And then you have at least realistic surface temperature, and then atmosphere which is consistent with that surface temperature, equilibrate ocean which is consistent with that surface temperature. But then you start running, instead of prescribing the surface temperature you take that constraint off. Then your model starts drifting from realistic surface temperature away, even though ocean was in equilibrium with that realistic surface temperature, atmosphere was in equilibrium. But once you start running, surface temperature started drifting away. Then sea ice disappears, snow disappears, all kinds of things happen. Then the model sensitivities are all wrong. So what I did was, when you start running from that, start drifting, then I ask myself in order to keep that temperature to be where it was, which is realistic, how much heat source and synch you have to put in. You ask to the model how much it is. So as model tries to deviate, "Oh, Syuki, I need this much heat here," to give it here, alright? Model say, "Will you go up? Syuki, I need this much negative heat here." Okay? So I registered on the computer how much heat source I think I need. Very systematic thing. And next time I run is I don't constrain the model one way or another, but I know how much heat source sinks I need at the surface in order for model to remain realistic. So put that thing in constant with time doesn't change with [???]. Once you make up your mind, once you start running coupled system, then it's an atmosphere with an exchange of heat. But you add that much in there, hoping by doing so model may not drift.

Edwards:

So then if you perturb it by doubling the CO2 or something like that then —

Manabe:

You have a same heat source.

Edwards:

— surface temperature will begin to rise, but it's still using that same correction.

Manabe:

That's right. So if correction is used for both perturbed and controlled experiment. So it's a fair game in that sense. And it's by so doing; suddenly I am able to do all kinds of interesting calculations. And this is what everybody else is complaining, because when they are working so hard to improve the model, here is a guy who came up with this gimmick, and without pain [?] due already using a model, all kinds of interesting results, which may or may not be true in their opinion. Okay? And then wants them to clean up the mess. He himself is creating a mess here. Right? That's what they think. And that's why they are angry, because — and instead of incorporating their parameterization, I am still using my old 30-year-old parameterizations in the atmosphere, and then using this ad hoc method, but I started having a great time, starting from latter half of 1980 starting all these calculations and started this Greenhouse Warming calculation by coupled model, which was my dream for a long time but I couldn't do it, because model is not realistic enough. Okay. And before that of course I was doing it some other way instead of having full ocean I have some kind of mixed layer ocean. One day ocean doesn't have any circulation, whereas before that I have a swamp ocean. So first 1979 paper of CO2 doubling I used ocean which is swamp, which is like real ocean in the sense it is wet. Right? But no ocean current or anything. And later I put in a, in order to put, if it's a swamp I tend to get too much summer-winter contrast, so I couldn't put the seasonal variation in. So you know I put the seasonal variation in, I put the mixed layer ocean, which gives you a pretty nice looking seasonal variations. And in this case ocean is not only wet but it has enough heat capacity to give you right seasonal variation of temperature. Which is very important to have in order to perturb that climate. So but then, so I was doing all these by using ocean model, which is very insulting to oceanographers. And so then only toward the latter half of 1980 I think a coupled model maybe, yeah, latter half of 1980 we started using coupled models with global computational domain and before that we are using coupled model with idealized domain like the other ones, you know, like this kind of thing, a sector computational domain we'd call it. And only latter half of 1980 we started using full global coupled model with flux adjustment. Then started doing all kinds of numerical experiments with flux adjustment but with putting Greenhouse gas in, and so that the IPCC 1990 is the result which presented in there is essentially our result, [???] etal [?], which Stouffer [?] is the man who worked with me, which we sent a paper to Nature 1989 or something like that. We kind of dominated the IPCC, first IPCC report because of that. And we found many, many interesting phenomenon.

Edwards:

Now when did other people also starting ocean atmosphere coupling?

Manabe:

Now ocean atmosphere coupling, early 1970. NCAR people were doing it. Larry Gates was doing it.

Edwards:

With CCM-0 [CCM-Zero?]? Did that have a coupled —

Manabe:

Yeah, CCM-0 coupled model. And Larry Gates was coupling Mintz-Arakawa's model into it. This is all after we did the 1969 paper. So when we are sort of developing this asynchronous coupling and all these things, they are doing it but they, I don't think they really used, they worried about this kind of thing. They coupled it then runned [?] it. So it's not much meat to it really in terms of not only in terms of coupling strategy. And our simulation wasn't good. They also present their results, their simulation. We too got this kind of result, you know, that kind of thing. And so essentially there were three groups who were doing it at that time. And we are leading basically the direct; we are setting a direction I think. This is maybe too much to say it myself, but that's what I think. And after early 1980, we started using idealized geography coupled ocean model and latter half of '80s we started doing [???]. And NCAR people actually in 1989 they published also transient response experiment by using coupled model. But with huge climate drift. So Northern Hemisphere is infested with sea ice all over the place, so Southern Hemisphere, all sea ice disappears rapidly. And so even after they put the CO2 Greenhouse gas in, model is drifting so that the entire globe is cooling down. So then control experiment is cooling even faster. They subtract to [two?] cooling down climate, and the difference is the Greenhouse Warming effect. So this shows that if you carry out a numerical experiment without a stable platform, which is a control experiment, if control experiment is drifting very rapidly you cannot rely on your perturbation experiment, because you are studying this climate change through, by trying to shoot moving target. And so that's why — And the 1990 IPCC report at that time Hamburg people start catching up and they started getting a result. But —

Edwards:

At the Max Planck Institute?

Manabe:

Yeah. And they have a brief mention of their work in the 1990 IPCC report. But they haven't published yet, result yet you know. So I guess we have safely beaten them at that time. And but in terms of analysis, because experiment of ours come out much cleaner, it's so much easier to analyze and that shows how important the control experiment, doing control experiment right. And 1995 report then now everybody using coupled model. I am really satisfied about it myself, although they are sort of looks like competing with each other, but essentially everybody doing the way I proposed, and some people who criticized about flux adjustment go ahead and use it still. So by criticizing it you know your conscience is satisfied. They want to use it! And I think it is a very good approach. However there is another approach one can do. Instead of admitting okay, we can't completely balance the heat balance here, you actually tune the parameterization until you get pretty nice balance. Okay? So this is a much harder way, because in my way you know you ask model how much you need and then you put like this so you can determine magnitude flux adjustment in no time. On the other hand if you try to parameterize cloud parameterization or convection parameterization like this and then try to do it, it's much more difficult, because so many parameters to adjust and —

Edwards:

And you just have to keep running it over and over and see what happens.

Manabe:

Over and over again, yeah. That's what the NCAR people are doing. So I guessed these parameters, but then even when you perfectly balance the imbalance given the realistic surface condition, still there is a possibility that you are over-tuning one component of model when other component of model is not satisfactory. So that people seem to choose tuning of the parameterization to flux adjustment, even though it is much harder to do that. They don't seem to realize just because they are able to tune them right so that drift is minimized doesn't mean their tuning is realistic. Right?

Edwards:

Yeah, I see.

Manabe:

And so that it's not as black and white [as] everybody preaching, "Oh, flux adjustment is evil. My tuning is right," you know. It's nothing like that. It's really an area of gray, and it's so much easier to do it in flux adjustment. And of course flux adjustment, the smaller it is it's better. So you keep track of how much smaller you have made every time you improve the model. Of course you have to do that. But on the other hand it is a very convenient technique where model is imperfect but still you want to get the result. See, this is what I meant is that I am willing to take shortcut in order to get the result, because I don't want to wait forever. And whatever shortcut I can take, I would like to take it. The other people say, "Oh no, this means parameterization is no good, so I have to go back to drawing board." Let them do that, but I rather want to go to my goal. And so, and the rate if result of experiment is incorrect, they will find out later. Okay, that's fine. But if I'm right, I get the credit for it.

Edwards:

Let me ask you about another thing that's sort of related to this.

Manabe:

Yeah.

Edwards:

In the last few years, and maybe before, one of the things you started to do was some very long runs of these coupled models, maybe thousand year runs.

Manabe:

Yeah, right, yeah.

Edwards:

What was the motivation for that?

Manabe:

The motivation is very — One of the things is, of course everybody running toward higher resolution, more sophisticated parameterization. I decided to run opposite directions. But that was sort of usually my reactions. That is, I told you at the beginning, you are coming from Japan, I am kind of relieved. And after that ever since I am running, trying to run the opposite direction from the pack. Because if we run with the pack, if I go there, I want to get that much benefit. I have to share with others. So I want to, I decided to go around the other way. Instead of going higher resolution I keep it low resolution and runned on time [?]. What how I found out is that because parameterization is simple — for example if I eliminate diurnal variation of sunshine I can run model four times as fast. Just eliminate diurnal radiation. Result is hardly different. I have tested. Yes. You see, there is many —

Edwards:

It's run with a constant —?

Manabe:

Yeah. It goes on constant sun during day and night, constant sun, climate hardly changed. I had done the homework, just as for other things. I had done the homework. I am not a lazy person. I just do it and then make sure it's okay. And then I do it that way, and if I can save computer by factor of four, don't you want do it? I do. And but what happened is, as I ran this model thousands of years, then I suddenly found out that the model is simulating many of the climate fluctuations which have been discovered, sometimes vaguely, sometimes clearly. Both Atlantic and the Pacific Oceans and surrounding regions like multi-decay of oscillations which looks like many times greater serenity than normally, the Northern Atlantic Ocean. Cold, fresh water comes gushing out of Arctic Ocean through East Greenland current to North Atlantic and creates a large sea surface temperature normally. Every time it's come out it's maybe several decades. That's a very interesting phenomenon. When it's come out sometimes very often it influences this conveyor belt circulation. That's also multi-decay of oscillations. And the model seemed to simulate that well. And also there is North Atlantic oscillations, which is essentially forced by atmosphere, and the ocean is responding at the old time scale. And these phenomenon are simulated very well. Also decay of the tropical oscillation in the Atlantic and in the Pacific decay there are oscillations of extra tropical decay, there are oscillations of sea surface temperature, tropical, extra tropical combined decays of oscillation. All these very long term fluctuations which people started seeing a glimpse of it is all started coming out in these models. Originally I was learning it because I wanted to learn long enough so that I want to see a background natural variability so that when you have a Greenhouse Warming, warmed earth gradually emerging from this natural variability, I was to know more clearly about hey, what is coming — which I think it has already started emerging out of natural variability. And so that's what I think. And so that in order to detect the Greenhouse Warming from natural variability, you've got to know the natural variability. And so in order to do that if you have a fluctuation of decadal [?] 50 at time scale, you have to run minimum of thousand or two thousand years in order to get reliable statistics. So that is why I'm learning long, long distance. And this is — and of course you know this coupled model, the thing which I haven't done much myself, I started doing recently, is El Nino southern oscillations. This is one of the important applications of coupled models which started emerging recently, and so that the entire climate modeling community now got into this coupled model study, both natural and manmade change, and in my opinion the golden age of coupled modeling [has] come about, and I am very pleased with it. And you will hear a lot of things about coupled model studies and these things, or detection of Global Warming, and coupled models [have] become a very powerful tool, not only studying natural variability but also Greenhouse Warming and then sort of detectability of this Greenhouse signal from the natural variability, and so that — And then now — And all these groups, Hudley [?] Center is working on it, Max Planck is working on it, NCAR is spending full energy onto it, and GFDL of course has been doing it. And then Los Alamos are going to get the super duper computer and to try to put some of the coupled models into it, and they are crying wolf about U.S.'s modeling gap from European and trying to appeal to the sort of you know insecurities they may create in the mind of funding officer. I guess that would be alright, rather than try to think you know we are ahead of everybody and we are safe and without working very hard, rather better to say we are behind and we have to do something. So I am very glad that they are doing it. But I think that climate modeling has get into sort of second stage. Now weather modeling I told you, Charney, Phillips first started with instigation of Von Noyman's and all these creative groups at the Institute for Advanced Study —

Edwards:

Right.

Manabe:

And then until the time of Von Noyman's death they went to MIT. That's sort the end of a sort of romantic age of weather forecasting. Okay. Now then of course GFDL of course one more thing happened. GFDL started using general circulation model which developed from weather forecasting model back to weather model, started demonstrating that weather forecasting can be extended by including various physical processes such as radiative transfer and more explicit hydrologic cycle we would be able to improve the weather forecasting, extend the weather forecasting. And that sort of led into European Center climate modeling. So then from the beginning they started putting all these parameterizations and various things in, and then tried to extend the weather forecasting. Now that's —

Edwards:

One of the things I was going to ask you about is connections that you've had with the Europeans. This would be a good time to hear about it.

Manabe:

I don't have much. I only had, Joe and I kind of provided Miakoda to do extended range forecasting, demonstrate the feasibility of extending the range of forecast. So through Miakoda we kind of instigated establishment of European Center. In that sense I think we are responsible [for a] significant part of it. You can ask European Center people. So we demonstrated the feasibility of extending by using GCM, which was more used for climate purpose, but you can use it for weather forecasting.

Edwards:

Last night I talked to somebody at the [???].

Manabe:

Yeah. This point I want to make, one more point is so what I am saying is that when Charney and Phillips and all these people were working until maybe GFDL was doing you can call romantic age of weather forecasting. Okay. But then you get into second stage, which is European Center. They start gaining good applied mathematicians; they started getting good super-duper computer, dedicated super-duper computer only for that purpose. Okay. And then they started doing a very systematic approach. They get various models and tune them, you know, in a way they can tune them, and then do weather forecasting in a very systematic way, increasing computational resolution — which should be good in order to express hydrodynamical equation. So it's getting to sort of leadership of organization man is required. Steve Jobs [?] is [???] too. Somebody else, another guy, next guy who went to somewhere else, Naviscore [?] or something there. But he was second guy.

Edwards:

John Scully [?].

Manabe:

Scully. Scully. So yeah, so Steve Jobs to Scully. Okay. That is the evolution of weather forecasting modeling. Climate modeling may be in that transition. Syuki Manabe may be a Steve Jobs, but may not be ideal person for getting into this Los Alamos or thing like that. And you may need more organizational men, more computer experts, more applied math, and so that you can get a maximum number of computation done. And then it's more regional climate change, 10 kilometer masses, and sort of you know much larger scale organizational thing is needed. Now okay, you say oh no no, climate modeling is not that, because we have to solve the parameterization of cumulus convection, we haven't solved the parameterization of land surface process, we haven't solved the cloud microphysics and cloud prediction problem. Of course I know. But the thing is, it takes a long time to solve that problem, for the reason I explained to you. Okay? So that it takes a long time. Meanwhile, since we can simulate the climate pretty well, at this stage we can go to that second stage and then get the regional climate change information. Why not? Or when you try to predict southern oscillation in [???] in El Nino, you can have a higher resolution so you can predict the southern oscillation more accurately, use these very high resolution models. If you use the same high resolution model, you may be able to represent interaction between El Nino and the United States climate, which is the subject of this winter so much. Right? And so the climate modeling is getting to be from romantic age to organizational man age. That's what I think.

Edwards:

If I were going to talk to two or three people at the European Center, who would you recommend? Especially who know about the beginnings?

Manabe:

Axelwin Needleson [?], he's a Denmark. Axelwin Needleson. You go to Axel. He is the one who established European Center. And Ed Benson [?]. And these are the two people I will go. And then if you want to know the latest development, of course current direct, what is her name, I've forgotten again. Anyway, you just ask director's name and Burridge, Burridge. B-u-r-r-i-d-g-e or something. Burridge. Okay. And these three people are three directors, and they all of them are very good salesmen, and so they are very articulate, eloquent in explaining what they have achieved. And so that's what I would do. Now then so much about this coupled model, but then the question then is I say oh it's a time for this organizational man, and you go to this very high resolution model and do best out of computation of hydrodynamical equation, and you got a regional climate. In order to get to regional climate you have to get all these storms, simulation of storm tracks and all these things right. That means you have to have a good weather forecasting model into climate model. That means a very fine resolution model, and to try to get the regional forecast. And so that's how the west coast of the United States specification is going to change, how it is going to change in Seattle, how it's going to change in Florida panhandles and golf course. You would like to know how the rainfall pattern will change associated with Global Warming, how the hurricanes are going to change, and all these things. Now it's currently intense debate. I have worked a long time on this myself, but it's intense debate. And so they are going to do all these things. However, there is one problem, because okay, you increase the resolutions, but then, and you get a regional climate change. But you have to remember that the current estimate of the climate sensitivity, Global Mean Temperature in response to doubling of carbon dioxide is between 1.5 to 4.5 with most likely value of 2.5. Now this 1.5 to 4.5 range has been there since 1979 when only Jim Hanson and I were doing some — Okay, Jim Hanson just got into this field. His first result came out '79. So he and I and they arbitrarily decided between 1.5 to 4.5. They haven't changed the direct means uncertainty hasn't increased, probably uncertainty has increased. Some people may want to increase the range even more, but they are afraid to increase because people then say, "How come you spend all this money for this" —?

Manabe:

Peculiar things are going to happen. That is, when you start getting a finer and finer scale projection of future climate change, but you don't know the global mean temperature change by factor of three, and maybe even more. Now why then, what's good about having regional climate change by super-duper computer when you don't even known Global Mean Temperature Change by a factor of 3. Isn't that emphasizing details first before you know the big picture? Which is a very bad thing. You go to trees rather than seeing forest. And so that is one of the things. And some people think by making parameterization of physics more sophisticated — more complicated in my opinion. Sophisticated is a very peculiar word.

Edwards:

[laughs] Yes.

Manabe:

And more complicated. Because nature is infinitely complicated, by making parameterization more complicated you approach toward the nature in complexity. However, no matter how complicated you make your model, nature always laughs at you because it's donkey with a carrot. Right. The smaller — you know the theory of factor [?] or something. The smaller you know you see more and more and more. Right? It's an endless thing. So by making it more complicated and competing with nature in complexity is similar to Don Quixote attempt. And this is not the way to do science anyway. And so then what do we do? And the thing I think that should be done is to, as the Global Warming sort of gradually come up — And you have a paper which I gave you earlier.

Edwards:

This one? [???] [???]?

Manabe:

Yeah. As the temperature comes out from a natural variability like this. This is model simulations and observe the thick line is to observe the temperature, and you can see that model is, if there is no Greenhouse gases or sulfate aerosols it will just continue like this indefinitely. However when you start putting forcing in there, then it starts warming up like this, which this one happened to simulate beautifully. This could be a lucky coincidence, which I could explain there. But as model comes out of natural climate variability and so that you can begin to see Global Warming yeah indeed take place now the middle of this century seems to go up. So what you do is you give the increase of Greenhouse gases and other thermal forcing of climate and then put that into the model. Okay. And then see how much the warming of the model is. Compare that with observed Global Warm of mean temperature, Global Mean Temperature, and then try to evaluate how well model is simulating observed warming. And even though this takes too long for many people's comfort, I don't see any other way. Because only way you can evaluate inaccuracy of the model, if you have a perfect model, okay, and you have your simplified — always model is a simplification of nature — and you have a simplified model, and then you compare perfect model projection with a simple model projection and see the difference between the two, you can see inaccuracy of that simple model. Unfortunately, there is no hope of creating perfect model which is this planet. Okay. So instead of creating perfect model, you look at the actual warming which is a perfect model; that is nature. Actual warming is nature's warming. Okay. Then you compare your model simulated warming and then make sure that model yields about right sensitivity. Now still this is not part of the solution. You may be getting right answer for wrong reason. But I think that is most — you do best you can in improving model, and every time you do you make a projection of observed warming and compare, and then make sure model is yielding, just as I did here. Now so that is — and then you make sure Global Warming is right ballpark, and then you put this super-duper computer and its runs a very high resolution model and get, fill in the regional detail. So that that will be useful for decision makers. Now unfortunately however one of the biggest answers [?] then is the climate forcing. Now we know what is the Greenhouse gas forcing, how it's increasing. With regard to aerosol, we know very little. I got this one from UK Met [?] office who pioneered putting aerosol into this thing. Before that we always over-shoot actual warming, and we put the sulfate aerosol, industrially produced sulfate aerosol, we happened to get perfect simulation, ever better than originally UK Met one here. And but there is many other aerosols other than sulfate.

Edwards:

Right. Sure.

Manabe:

Yeah. And then we also don't know exactly, so our constant is constant or not. We also don't know how much tropospheric ozone has increased. And so that we do not know the forcing of climate accurately enough to evaluate model sensitivity. And therefore here again instead of using satellites for so many wide varieties of problems, I propose to send the satellite in order to determine things like aerosol forcing. Because sulfate aerosol washed out by rain so that it has a wide spatial variability. So the only way to see the radiative effect of sulfate aerosol is satellite. And so that —

Edwards:

Are there not any instruments now that observe sulfates from satellites?

Manabe:

I think they could, but you know there is, they can put lighters [?] or something, and also they can — Some astronaut went up there and looked at lighter [???] and then scanned the earth from space shuttle, and that itself is not much of measurement, but it at least demonstrates the feasibility if you put such a thing on satellite they can do that. And so that I think still what and how we measure them is still a big unsolved problem, because there is the other kind of natural aerosol like Sahara dust, and all kinds of complexity. So natural variability of dust loading in the atmosphere is huge. So how you identify the man's impact [is] not easy. And so it is easy to say what I said, but no matter what you do unless you know the forcing of climate, how you can even, you can't even begin to project in the future reliably, or you can't of course evaluate a — So this is a fast [???] conditions in order to evaluate uncertainty of climate change.

Edwards:

Maybe this would be a good place to start talking about the IPCC.

Manabe:

Yeah. Now already I got in there, you noticed, this the problem of uncertainty. This is an issue. So yeah, let's go to IPCC now. Yeah.

Edwards:

A lot of people date the beginnings of the sort of movement to the Rio Treaty and IPCC and so on from the Vilaj Conference in 1985.

Manabe:

Yeah.

Edwards:

Were you at that? Did you —?

Manabe:

Vilaj. I was there. Yeah.

Edwards:

Did that strike you as a major event? Did it seem like a turning point in the history of your community or —?

Manabe:

No, wait a while. Maybe that Vilaj, there is maybe, I may be in Aria Vilaj, but I don't know. We have a Vilaj Conference there. And yeah, I think that what it is is that IPCC is, or Rio Conference, or things like that, put this debate of Greenhouse Warming not only scientific issue but also become very important political issue. And so that climate modeling and its application to projection of Global Warming is no longer the nice past time which I have done most of my career, just have fun with it. And the debate is not just scientific curiosity, so that every paper you publish [is] subject to severe criticism. And now there is always scientific criticism of your work, but ordinarily, "Oh, he published interesting work. I'm not sure he's right, but it's an interesting possibility and we have to evaluate this future, enjoy evaluating that possibility." That's a criticism, most of the people's reaction. But not anymore. You know, you publish something, "Oh my God, he is claiming that El Nino southern oscillation is going to amplify tremendously." Which is contrary to the result of my work, but such a paper may be published very soon in Nature. Oh, if it's published then you know big debate and — And because I think there is the people who feel that we should not prematurely around the society in this thing, and we shouldn't prematurely getting into mitigating actions when we don't know enough about it and its impact we shouldn't do it. So that they — and some of their concern is probably right, that when we don't know enough why do we spend so much effort. And so that's the thing, so that everything we do and say now becomes very much of the intense scrutiny, and every time it's getting out through IPCC process.

Edwards:

Yeah. Now how did you come to be involved in the IPCC? I mean you're an obvious person to be in it, but I wasn't sure —

Manabe:

Yeah. So the first time is that as I told you I avoid these committees like pests. So actually they came to Princeton to first meeting of IPCC's — Yeah. They say, "Oh, if you are not coming, we all come to Princeton," so they came to —

Edwards:

[laughs] Who organized that?

Manabe:

John Horton is the one, Sir John is the one at that time, and then the entire [???] Met office people and all these people came there, IPCC. They have that Aria IPCC meeting at Princeton. And then later they told me I better come to [it]. So I get all kinds of pressure from NOAA hierarchy I have to go to [it]. And next one, some kind of a northern England that is a, what is that? Northern England. Scotland city, capital of Scotland?

Edwards:

Edinburgh [?]?

Manabe:

Edinburgh, yeah. They have their meeting in Edinburgh. And that was a very crucial meeting. I didn't know. But then I got all kinds of pressure. I went there, and it turned out that it was the most important section of climate, transient response of climate to Greenhouse forcings. Section was not written because of Aria's situation reason, which I don't want to elaborate there, and so then they hired Francis Bretherton , who is an excellent, articulate writer, to write that section.

Edwards:

Francis?

Manabe:

Bretherton. Bretherton.

Edwards:

B-r-e-t-h-e-r-t-o-n?

Manabe:

Yeah. He is a very articulate guy. And so I helped him to, I gave a talk there explaining our latest result from coupled model, and Francis wrote a very articulate summary of that in IPCC Report #1. So transient response section relied very much on our results. Now the IPCC Report #2, the UK Met office came up with this aerosol forcing, and so that aerosol —

Edwards:

IPCC '92?

Manabe:

'95.

Edwards:

'95.

Manabe:

Five, yeah. In between one is an interim report, so '95. And they came up with this, starting already from '92 report that they begin to realize maybe the reason our model tended to overshoot, exaggerate warming during 20th century, may be due to aerosols such as sulfate aerosols. And that I think is pioneered by UK Met office people, although aerosol forcing we don't know enough. But if I put their forcing in, our model too give a very nice result, and so that it's no longer the monopoly of GFDL model anymore. Which is probably healthy. And I am kind of happy, because we tried to help UK Met office people send some of our people to do flux adjustment right and get initial condition right, we tried to help them getting their result. John Mitchell [?], who is coming here, is a very good friend of ours, so we do everything to help. I don't give a damn whether it's a U.S. group or foreign group. If I can help them, I will help them. So as a reason why we didn't spend so much time on the parameterization, the part of this thing is this coupling became a very, very difficult problem. We couldn't afford to do anything else. [Manabe coughing a lot...] [???] [???] drink this one. So what I think is, getting back to previous thing I was talking about, how to calibrate sensitivity of model, that every time we meet in IPCC we will have this problem, that is we don't know enough about the reliable, we don't have a reliable estimate of sensitivity. And I think that it is important for us to, rather than try to study all the problems of climate process, but to try to focus more on how to narrow down this uncertainty and by doing the thing which I just described so that decision maker will have a easier time on this. Now the coupled model application which I have discussed so far is essentially confined to the, given the increase of Greenhouse gases how future climate changes are going to be. However, we have started using same coupled model for estimating carbon cycle change into future so that right now we have just wrote, I'm a junior author on this [???] Sarmiento of Princeton University is the senior author.

Edwards:

Sarmiento, that's S-a-r-m-i-e-n-t-o?

Manabe:

Yeah. And application of same coupled model for the study of carbon cycle. And this is going to be another very important application which is going to grow very much. And what it is, is if you look at the Kyoto meeting this time you will find that most of the things they are discussing is where to stabilize carbon dioxide. People are not so much talking about climate change this time; most people are discussing how much they, what kind of a sort of target of constraining, reducing CO2 emission compared with 1990 level, how much [???] do it. And so that in order to get reliable estimate of that, then it has become very important to use these models if we can to try to let reliable projection of future level of CO2 and the scenario of future increase if we do what how CO2 is going to increase. And so you give the study of emission, future, if we do it this much you do get this, and that, and so forth. And so we have to use that for that purpose. Now unfortunately there is another the most difficult problem of this earth science which is to estimate how much, the way we are using so far is if you give the atmospheric CO2 increase in the future, what is the oceanic uptake of CO2. But more fundamentally, we would like to use this same coupled model to give future study of emission of CO2 by burning of coal and so forth, and then how many percentage of that is airborne and how many percent of that go to ocean, how many percentage of that is absorbed by land biosphere. Okay? So that then, in order to do that, in addition to coupled ocean atmosphere model in which we have a carbon cycle in the ocean model, put the carbon cycle in the ocean model, but also we need a model of biosphere which how much carbon dioxide is absorbed by land biosphere, how much would be emitted by oxidization of organic material, and so that net sinks or so-so carbon dioxide into natural biosphere. And that we have to put that into model, which is a very difficult problem.

Edwards:

Yeah. A lot of people began working on this in the early 1990s, this introduction of other, especially [???] of biosphere and carbon cycle and so on.

Manabe:

Right.

Edwards:

Was that an outgrowth of the IPCC work or [???]?

Manabe:

Oh, no. Before that people like Hans Eskar [?] of the Citron [?] and Segon Thorax [?] and all these people are working very hard on it, starting from the '80s and using much simpler, much simpler model — box model of ocean, box model of atmosphere, and then try to see, estimate, see how much airborne fraction of CO2, how much atmospheric CO2 has been changing, how much it's going into ocean, try to figure out some various mathematical techniques to figure out how much it's absorbed by land surface. So land absorption is usually used as a residue, residual. And or some people use atmospheric GCM to try to study where land is absorbing or ocean may be absorbing. And then sort of last ten years or so, when people keep adding up and subtracting how much is going into the atmosphere, how much going to land, they begin to find there is big holes in this argument [?] [???], and that's called a missing sink. And a lot of people think that is absorbed by land biosphere. And that could be estimated if you have a right atmospheric GCM, and you try to see a, try to have a, let atmospheric GCM move carbon dioxide around. But in order to maintain, observe the level of CO2 at earth's surface, where you need to sow [?] some sink, just like flux adjustment, you can figure out what [???] sinks do you need. Okay. And so you can put this thing, where is the missing sink? And that's what people are doing. So this is another application of atmospheric GCM, in order to determine what is the net source on the sinks on the surface of earth. Okay. And or more fundamental projection, instead of using box model, if you can have a land biosphere model of some kind — which I think is very difficult to be quantitatively correct — but put that in, and then try to make a future projection, rather than using simple box model. And that is going to become very important, because for example in this IPCC they have decided that they allocate emission right to each country. And so if U.S. for example helps developing country reduce the carbon emissions, then U.S. gets the credit for it. Well that's fine, that you can calculate how much emission [???] U.S., how much the U.S. prevented CO2 to be emitted? However there is another item in that negotiation. That is to try to estimate how the natural biosphere of each country is absorbing or emitting CO2. Oh, we don't have any way to measure. Now how you can negotiate putting an international treaty with something which there is no way of evaluating, measuring? That is a big mystery to us. Right? So one way to do it is give atmospheric GCM and you give the cheap, relatively mass-produced CO2 concentration measurement and do it all over the world. Okay. Weather stations all over the world, they measure CO2. Right now Hawaii and several stations in the world. But do it all over the world. And then move things, or move CO2 around by atmospheric GCM, and then try to determine the net source on sinks. If model is realistic enough, you will be able to determine, in order to maintain, keep that observed distribution of concentration of CO2 at very geographical locations, what source on sink do you need from the surface? You may be able to estimate which country is emitting how much. And that's what they are doing at Princeton, is [???] Sarmiento is doing at Princeton, and so this is the new kind of — So that in order to do this kind of thing, you need to do some of these calculations by using a model. And try to see how — The important thing is what natural biosphere is doing, and because modeling of, behavior of biosphere is so much more difficult than modeling physical system. Right? Therefore, we need instead of letting these people who parameterize anything, parameterize land absorptions. You know, if you tell them to, they will parameterize anything. Okay?

Edwards:

[laughs] Right.

Manabe:

If you give them money. And come up with a parameterization. We don't know how to do it right yet. Okay? But most of the current modelers parameterize anything if they have to. They invent if statement and some kind of algorithm. So I tell them, "If you can do all these things which parameterize something they don't fully understand, why don't they parameterize God?" They get angry. And so anyway, but I think to let them parameterize a single handed way, single-minded way, is not good by itself. There is always something to check parameterizations. And [in] most cases we don't even know how to check all these parameterizations. So they create very complication parameterization which [we] have no way of checking, or checking [in a] very indirect way. No way the rigorous check of parameterization. And then they claim my parameterization is very good or all that kind of thing. And so there is a danger to letting them do one-way street, and I think there is always some way to at the same time check their parameterization by this kind of somehow try to measure the emission of source on sinks. See whether they are doing right.

Edwards:

Yeah. I want to go back to the IPCC a bit. I am interested in your impressions of it as an organization, because it's an unusual organization in that national government representatives and NGOs are both able to participate in this.

Manabe:

Yeah. Right.

Edwards:

What are your impressions of its meetings and how the political forces interact with the scientific community?

Manabe:

Now you have to realize that IPCC itself is an intergovernmental program of climate change, and its mission is to provide the information for decision makers. Its mission is not to recommend them what to do. And this is, this misunderstanding is causing a great deal of problem for IPCC directors, people who direct, lead the IPCC, because people think IPCC is trying to influence political process, which they are not. Their job is not to recommend anything. And therefore what they have to do is to come up with a most reliable assessment and balance the assessment of the situation and let political decision makers to go from there. Because the bigger uncertainty — although what IPCC is evaluating is highly uncertain, but between what IPCC provides and what decision maker does, there is even bigger uncertainty.

Edwards:

Yeah, sure.

Manabe:

And so that the important thing is the realization that IPCC is not recommending, and we have to stay in science rather than tell them what to do. Because we are inviting trouble by telling them what to do. Therefore, IPCC should stay out of politics. I think has been a most important point. I think Bat Boreen [?], who is director of entire IPCC came yesterday or the day before yesterday here, a couple of days ago, and I kept making a mistake of IPCC recommend, and he said every time I slip my tongue and said it he stopped it [???]. And so this is I think is a very important thing, and but I do feel that IPCC to be effective in future, every time they meet, every five years, the [???] five years, they better — I told him, although he didn't agree, but in order to reduce uncertainty next time both in carbon cycle problem and the Global Warming problem, that we have to do certain things, such as monitoring aerosols or things like that, or sort of know more about missing sinks next time we meet together, and things like that. Because these are the crucial issues which [are] obscuring, which it make it very difficult for decision makers to make up their mind. Which way they make up their mind, that is not our business, but at least we are responsible to [do] the best we can. Now IPCC is assessing a current state over a knowledge; however, I wish IPCC also decides what we can do in order to reduce the uncertainty, both in carbon cycle and Global Warming. But Boreen still thinks it is not his business. Maybe so. And it's what IPCC thinks they — they have their opinion what they would like to know if they, in order to make more reliable assessment. However it's not conveyed to world climate research programs or IGBP scientific body. Scientific body is doing not so much reducing uncertainty. Study all the problem of relevant to climate change. And the IPCC could influence when they prioritize — some with observation from satellite, or some of the scientific problems, what kind of funding, where funding should go, rather than just a chunk of funding should go. And that should be influenced by IPCC desire to know what sort of thing they had to know in order to achieve the desirable assessment. And that I think, personally I think that should be done, but current clivar [?], climate variability called clivar and World Climate Research Program, and join scientific committees and all the subcommittees they create tended to be more interested in natural climate variability rather than anthropogenic climate change. And for the detection of anthropogenic climate change, the study of natural variability is very important, as I emphasized earlier. But I think that they have, they need to have [a] more focused, forceful program of anthropogenic climate change modeling, as well as observations. Which I emphasized earlier. And that's I think what we need in order to — Every time they meet they get sort of exhaustive meeting, and each person have a different idea and they seem to be making decisions out of exhaustion. And now I'm not sure whether that exhaustion will be reduced for us providing more reliable information. That I am not sure.

Edwards:

Yeah. It's a big, big [???].

Manabe:

Because, yeah, a big project, and I mentioned earlier that one country's benefit may be a negative thing for another country. One group's benefit may be another group's dis-benefit. Is there such a word?

Edwards:

Burden. That's the opposite thing.

Manabe:

Burden. Burden. Yeah. So that's what I think. And also the fact that as I mentioned earlier, that technology development such as biotechnology will enable us to adapt if a warming occurs slow enough. Right? So many people may think that why should we suffer now when we think can adapt it later. And that's why. So the reason why they get exhausted [is] not just the uncertainty of our projection, but there is many other reasons, societal reasons and everything. And that's why it is so difficult to reach any consensus, what we should do about it. And so that is a problem. But on the other hand I think as a scientist we have a duty to try to reduce this uncertainty to provide them the most reliable possible information.

Edwards:

Well, speaking of that, so what's your view of the quality of the IPCC reports that have been produced so far?

Manabe:

I think it's pretty well balanced. I am rather impressed by the very careful way they go about being reviewed, even though it is impossible to satisfy everybody. And the way they are written, I think it's very balanced, well balanced, and if I — I am thinking of writing it singlehandedly myself. I don't know how better I could write in terms of — Of course I would write more, because if I write myself, I would put my dogmas in, or result, dogma means result of my own work rather than everybody else. But so that in that sense they are doing better than I would write it myself. They are very well balanced, and people criticize the consistency between thousands of pages of the basic document and policy makers' summary, but I don't know how you can extract information in a perfectly logically consistent way to this policy makers' summary. There is here and there some seeming apparent inconsistency or some kind of intuitive jump from one to another. When that happens everybody screams. But I think that is unfair to them if they have to do it themselves. And strictly speaking, what is mistake is the lone [?] perceptions about IPCC report. Because people think that IPCC report is a consensus report. Now the word "consensus," what does it really mean is the question. Consensus is in this case consensus means is the opinion which majority of people think that this is really what it should be. But everybody does not necessarily agree. But weighs pros and cons and they come up with this assessment. And therefore —

Edwards:

So you think it's more like a majority rule than consensus that everyone agrees.

Manabe:

Yeah. And with some subjective intuitive judgment of the various possibilities, and based upon careful scientific evaluations, all of the people who hold their pen so that it may not necessarily even — I think it's usually majority opinion, but it's not you adding up everything, or you say you only identify those opinions which everybody agrees. Now if they do that, probably little meaning of consensus is that. If they do that what they can do is very easy: they put beautiful covers, right? But blank sheet of paper between them. Right? They put blank sheet of paper between them. Because you know people never, never agree. You can't prove everything 100 percent. And that's why OJ got acquitted. Right?

Edwards:

Right.

Manabe:

And in order to sort of decide somebody is guilty in the criminal trial you have to prove 99.999 percent practically. If you have a videotape of OJ, people say, "Maybe they switched the face." Okay. So there is no way. And there are people who don't believe in Continental Drift, there are people who don't believe in evolutionary theory, some respectable scientists don't believe in it. And so in order to prove consensus in the real sense of the word, you will never get. And therefore you are daydreaming if you think that this is a consensus report. It is not simply.

Edwards:

Okay. Well, I think we should start trying to wind up here.

Manabe:

Yeah. Yeah, okay, yeah.

Edwards:

And I'm thinking about my last question, which is you know what do you think your three or four most important contributions have been. I mean your career has been so long, you've done so much that it's probably hard to boil it down to just a few things, but I'd like to hear what you have to say about that.

Manabe:

Yeah. Now the most important contribution is probably doing study of Greenhouse Warming by using model, particularly I think when they have encountered difficulty about getting the, using the method pioneered by Arenius. And then they are going nowhere. I was able to use my one-dimensional radiative convective models and try to include the effect of convections and then get more physically based estimate of Greenhouse Warming, propose a methodology to do it right. And that I think [is] probably the most important contribution. And then once that happened, then after that I use more and more sort of —

Manabe:

...the Greenhouse gas increase induced change, the physical mechanism that controls the climate change, I tried to sort of smoke out more and more as time goes on, which instigated others to do the same. And so that's probably the most important I've done. And another thing is I think is in the terms of GCM, general circulation model results is of course I was very fortunate to be able to work with Joe Smagorinsky. But I think what I did is to try to change basically dynamics dominated GCM into putting hydrologic cycles, atmospheric cumulus convection parameterization which I think, in my opinion, works very well for this purpose. It doesn't work too well in the weather forecasting for the reason I could I explain, but anyway, it worked very well so that it made putting hydrologic cycle into the model in collaboration with Joe, and radiative transfer so that explicit interaction between radiation and dynamics, explicit interaction between hydrologic cycle, radiation and dynamics, and then create some model of climate there and then try to extend that model and evaluate and demonstrate that model can simulate marvelously the general circulation of topics which [are] considered to be very difficult at that time. And then also are able to simulate the global seasonal variation of hydrologic cycle very successfully, early on before anybody else able to demonstrate following the work of, you know, first collaborate with Joe. But I keep pushing and demonstrate that despite extremely simple parameterization of process we are able to produce pretty realistic climate and general circulations. And so that sort of paves the way for later studies of [???], El Nino and all these stuff. And I think people usually don't give me credit, but one of the reasons golden age of El Nino research is stemmed from that very successful simulation of tropical general circulation which I achieved around 1970 before anybody else. And so that is another thing I think. And then of course using that model also we are able to simulate the stratospheric tropospheric interactions, the stratospheres. So the hydrologic cycle, tropical general circulation stratosphere we achieved landmark simulations early on, demonstrate that we can reproduce these things [with] relatively simple parameterizations. So that's the sort of achievement number two. And then another one is which I did in collaboration with Bryan, and to start developing coupled ocean atmosphere model which are now being used by practically everybody, developed all over the world and being used by both for the study of anthropogenic change and natural variability of climate change. And it's proven to be a very, very powerful tool, despite all the criticism people level against me about flux adjustment and this and that, I think that these coupled model really instigated a large number of studies since that time. So I think that these are the more or less the three — so coupled models, its application to natural manmade changing climate studies. And I guess these three are more or less the summary of three achievements. And then there's probably have all these papers there, but I can probably classify one of these things into these three.

Edwards:

Okay. Any last things you want to say for this tape?

Manabe:

Not much. I guess I had it enough. [laughs]

Edwards:

Okay. Well thank you very, very much.

Manabe:

Yeah, yeah, yeah. Thank you for listening [to] this. [laughs]

Edwards:

Very helpful for me.

Session I | Session II